Download Research Centre for International Economics Working Paper: 2013046

Research Centre for International Economics
Working Paper: 2013046
Title of Paper
Why Do Firms Evade Taxes?
The Role of Information Sharing and Financial Sector Outreach
Authors’ List
Thorsten Beck, Tilburg University and CEPR
Chen Lin, University of Hong Kong
Yue Ma, City University of Hong Kong
Abstract
Tax evasion is a widespread phenomenon across the globe and even an important factor
in the ongoing sovereign debt crisis. We show that firms in countries with better credit
information- sharing systems and higher branch penetration evade taxes to a lesser
degree. This effect is stronger for smaller firms, firms in smaller cities and towns, firms
in industries relying more on external financing, and firms in industries and countries
with greater growth potential. This effect is robust to instrumental variable analysis,
controlling for firm fixed effects in a smaller panel data set of countries, and many other
robustness tests.
© 2013 by Yue Ma. All rights reserved. Short sections of text, not to exceed two paragraphs, may be
quoted without explicit permission provided that full credit, including © notice, is given to the source.
Why Do Firms Evade Taxes? The Role of Information
Sharing and Financial Sector Outreach
THORSTEN BECK, CHEN LIN, and YUE MA*
ABSTRACT
Tax evasion is a widespread phenomenon across the globe and even an important factor in the
ongoing sovereign debt crisis. We show that firms in countries with better credit informationsharing systems and higher branch penetration evade taxes to a lesser degree. This effect is
stronger for smaller firms, firms in smaller cities and towns, firms in industries relying more on
external financing, and firms in industries and countries with greater growth potential. This
effect is robust to instrumental variable analysis, controlling for firm fixed effects in a smaller
panel data set of countries, and many other robustness tests.
_____________________________________________________
* Beck is with Cass Business School, Tilburg University and CEPR, Lin is with the Faculty of Business and
Economics at the University of Hong Kong, and Ma is with City University of Hong Kong.
We thank Campbell
Harvey (the Editor), an Associate Editor, and three anonymous referees for their very constructive and helpful
comments. We also thank Jacques Mélitz, Alex Popov, Vojislav Maksimovic, Steven Ongena, Yuhai Xuan, and
seminar participants at Tilburg University, Heriot-Watt University, University of St. Andrews, the Financial
Intermediation Research Society (FIRS) Conference 2010 in Florence, and the Annual Bank Conference on
Development Economics (ABCDE) 2012 at the World Bank for their helpful comments. We thank Pennie Wong,
Kuo Zhang and Xiaofeng Zhao for excellent research assistance. Lin gratefully acknowledges the financial support
from the Research Grants Council (RGC) of Hong Kong (Project No. T31/717/12R).
1
A growing literature dating back to King and Levine (1993) demonstrates the important
connections between financial development and growth, a link that has spurred further
exploration into various aspects of financial development and their influences on the economy. 1
Much of this evidence, however, relies on macro country-level data and aggregate measures of
financial development (e.g., private credit to GDP ratio) and economic performance (e.g., GDP
growth). Using a large sample of firm-level survey data on more than 64,000 firms across 102
countries over the period 2002 to 2010, this paper assesses the relationship between financial
outreach (i.e., credit information sharing and banks’ branch network penetration) and the
incidence and extent of tax evasion.
Compared with existing literature, our study focuses on a specific dimension of financial
development - financial sector outreach - and an important economic outcome at the micro level
- firm tax evasion. Moreover, the nature of the database we use allows us to exploit crosssectional variation across firms within a country and time-series variation on a subsample of
firms with panel data available. This helps alleviate concerns of omitted variable bias and reverse
causation. The study therefore adds to our understanding about how financial systems help
reduce information asymmetries and increase the opportunity cost of tax evasion.
Studying tax evasion is important for both academics and policy makers alike. First, the
degree to which financial market development influences the public policy space, that is, the
fiscal, monetary, and exchange rate policy options, has gained increasing importance. 2 The
ability to collect taxes is an important though often underestimated dimension of fiscal policy.
According to Schneider and Ernste (2000), tax evasion is a widespread phenomenon across the
developing (and even the developed) world, with estimates of tax evasion above 50% in many
low-income countries. 3 More importantly, tax evasion has been viewed as an important factor in
the current sovereign debt crisis as it creates fiscal instability and deficit. According to a member
of Greece’s central bank, the extent of tax evasion in the country is about one third of its total tax
revenue, or about the same size as the country’s budget deficit. 4 There is a concern that tax
evasion may also drive other European countries such as Italy into crisis. 5 Identifying policy
2
areas that have a first-order effect on reducing tax evasion is therefore critical for policy makers.
Second, tax evasion has repercussions for information asymmetries, corporate
governance, and agency problems among a firm’s stakeholders. As pointed out in the literature
(e.g., Desai and Dharmapala (2006), Chen et al. (2010)), tax evasion creates opacity and
obfuscation, which could be used to mask rent diversion activities such as earnings management,
related party transactions, and other opportunistic behaviors by management. Kim, Li and Zhang
(2011) find that tax avoidance facilitates rent extraction by management and information
hoarding activities, and as a consequence increases the risk of stock prices crashing. Therefore,
understanding how specific policy dimensions in the financial sector can help reduce information
asymmetries and agency conflicts has been a first-order issue for financial economists.
Existing literature suggests several channels through which banking sector outreach, that
is, better information sharing and branch network expansion, might affect the benefits and costs
of corporate tax evasion. First, higher banking sector outreach increases the opportunity costs of
tax evasion by raising the likelihood and benefits of gaining access to formal finance. Beck,
Demirguc-Kunt, and Martinez Peria (2007) find that banking sector outreach helps reduce firms’
financing obstacles. Furthermore, as documented in the recent literature, credit information
sharing is associated with lower transaction costs (Miller (2003)), improved availability and
lower cost of credit (Brown, Jappelli, and Pagano (2009)), lower levels of lending corruption
(Barth et al. (2009)) and lower levels of bank risk taking (Houston et al. (2010)). In more general
terms, Johnson et al. (2000) point out that firms are more likely to hide output in economies with
underdeveloped market-support institutions because they gain little from being formal. Overall,
this would imply higher benefits from correctly reporting firms’ sales in economies with more
effective credit information sharing and higher branch penetration by gaining access to the
formal financial sector.
Second, more effective information sharing and more extensive branch penetration
reduce information asymmetries and agency problems between lenders and borrowers and thus
decrease the benefits of tax evasion. To evade taxes, firms inevitably need to manipulate their
3
financial information (“cook the books”). As documented in the literature, firms suffer
significant reputation losses and incur much higher financing costs due to illegal misconduct
such as corporate misreporting (e.g., Graham, Li, and Qiu (2008)). From a bank’s perspective,
tax evasion signals low quality of disclosed company information and other aspects of the firm's
operations. 6 Tax evasion also raises bank’s concerns about the firm’s future prospects and
default risk because tax evasion is usually associated with significant legal liabilities. 7 As a result,
information asymmetries and agency problems between borrowers and lenders increase with tax
evasion, which in turn affects banks’ lending decisions and requires banks to monitor firms more
intensively (Lin et al. (2011)). The higher costs are passed along to borrowers in the form of
reduced credit availability, higher interest rates, and more stringent loan terms (Graham, Li, and
Qiu (2008)). In an economy with higher branch penetration and better credit information sharing,
information related to corporate misconduct can be more easily observed and shared among all
other potential lenders, which will make it more difficult and/or more expensive to receive future
loans (Jappelli and Pagano (2002)). 8 Hence, the opportunity costs of engaging in tax evasion
should be higher in countries with higher branch penetration and better credit informationsharing mechanisms.
However, there may be countervailing effects due to lower cost of funding and less
reliance on collateral in financially more developed economies. First, better developed credit
registries and higher branch penetration might reduce average borrowing costs and thus make the
potential increase in financing costs due to tax evasion more affordable to firms. 9 Another
possible countervailing effect may stem from collateral being less important to creditors because
the information gap between creditors and borrowers is smaller and because creditors can
monitor firms more effectively in economies with better financial outreach. 10 Therefore, from
this perspective the overall opportunity costs of tax evasion may be either higher or lower in
more financially developed countries, which leaves the question for our empirical tests.
Using a unique data set across 102 countries and over 64,000 firms, we examine the
relationship between branch penetration, credit information sharing, and corporate tax evasion.
4
We find very strong evidence that credit information sharing and branch penetration are
significantly and negatively associated with the incidence and extent of tax evasion, suggesting
that the net effect of banking sector outreach on corporate tax evasion tends to be negative and
significant. This result is robust to controlling for a standard indicator of financial depth and for
an array of other indicators of the institutional and taxation framework of the country in which
firms operate. To address potential omitted variable biases, which might bias the empirical
results, we confirm our findings using an instrumental variable approach, using a subsample with
sample weights, and using an array of other robustness tests.
Using the same analytical framework as above, we conjecture that the relative benefits
and costs of access to formal financial services vary across firms of different sizes as well as
locations. Smaller firms and firms in smaller cities and towns stand to benefit more from gaining
access to formal financing than larger firms and firms closer to the economic center of a
country. 11 Similarly, firms that depend more on external finance for technological reasons, such
as a long gestation period or indivisibility of investment, as well as firms with higher growth
opportunities benefit more from access to formal finance than others (Rajan and Zingales (1998),
Houston et al. (2010), Bekaert et al. (2007)). We should therefore observe that credit information
sharing and branch penetration have a stronger relationship with tax evasion for smaller firms,
firms in smaller towns, and firms that rely more on external finance and have higher growth
opportunities. Our empirical results strongly confirm these hypotheses. The relationship between
credit information sharing, branch penetration, and corporate tax evasion is indeed stronger for
smaller firms, firms located in smaller cities and towns, and firms in industries more dependent
on external finance and in industries and countries with better growth prospects. Showing the
differential effects of credit information sharing and branch penetration across firms of different
characteristics, combined with the instrumental variable approach, helps alleviate concerns
stemming from endogeneity and omitted variables.
Finally, exploiting within-firm variation over time, we confirm our results for a more
limited panel sample of 3,800 firms across 42 countries, many of which introduced credit
5
registries or upgraded them in the early 2000s and have seen changes in branch penetration.
These firms were interviewed in two survey waves so that we can directly observe whether there
is a relationship between changes in the quality of credit information sharing, branch penetration,
and firms’ tax evasion. We confirm our results both for the level and the differential effect of
credit information sharing on tax evasion, further alleviating endogeneity concerns.
Our paper bridges and connects several literatures. First, we add to the literature on
finance and growth by exploring the link between specific dimensions of financial sector
development and the incidence and extent of tax evasion. The empirical findings shed light on an
important channel through which financial intermediary development can improve economic
growth and increase the fiscal policy space. 12
Second, this paper is related to the literature on tax avoidance and agency problems
within the firm. Dyck and Zingales (2004) and Desai, Dyck, and Zingales (2007) find large
cross-country variation in private benefits of control, related to – among other things – tax
compliance. This literature, however, focuses mostly on conflicts between management and
block shareholders, on the one hand, and minority shareholders, on the other hand, and the
importance of corporate governance in reducing managerial rent extraction. 13 Our paper relates
to agency problems between lenders and firms, with tax evasion exacerbating this agency
problem, but less so in countries with better information sharing and higher branch penetration.
While previous literature uses measures of tax avoidance and shelters based on financial
statements and/or stock prices (mostly for the U.S.), we focus on a much cruder phenomenon and
more important challenge in developing countries, namely, underreporting of sales and income
or tax evasion. 14
Third, this paper is related to a small but growing literature on credit information sharing.
Building on theoretical work (Pagano and Jappelli (1993), Padilla and Pagano (1997)), Djankov,
McLiesh, and Shleifer (2007) find that both creditor protections and information-sharing
institutions are associated with higher ratios of private credit to GDP using country-level data in
129 countries. Moreover, credit information sharing is associated with improved availability and
6
lower cost of credit (Brown, Jappelli and Pagano (2009)), lower levels of lending corruption
(Barth et al. (2009)), and lower levels of bank risk taking (Houston et al. (2010)). More recently,
based on a randomized field experiment, Gine, Goldberg, and Yang (2012) explore the
importance of personal identification for credit market efficiency. Our paper adds to the literature
by finding evidence that credit information sharing is also an effective device in curbing
corporate tax evasion.
The remainder of the paper is organized as follows. Section I describes data and
methodology. Section II discusses our results. Section III concludes.
I. Data and methodology
To test the relationship between banking sector outreach and the pervasiveness of tax
evasion, we combine firm-level data from the World Bank-IFC Enterprise Surveys with
indicators of financial sector depth, breadth, and infrastructure as well as other macroeconomic
indicators. This section discusses the different data sources and variables that we use and our
methodology. In this context, we focus on how to best control for endogeneity and omitted
variable biases that might drive the relationship between banking sector outreach and tax evasion.
The Appendix provides definitions and sources for all variables.
A. Data
We use data from the World Bank-IFC Enterprise Surveys to measure the degree of tax
evasion and to construct an array of firm-level control variables. The Enterprise Surveys have
been conducted over the past 10 years in over 100 countries. 15 Standardized survey instruments
and a uniform sampling methodology have been used to minimize measurement error and to
yield data that are comparable across economies in the world. The surveys try to capture business
perceptions of the most important obstacles to enterprise operation and growth, but also include
detailed information on companies’ management and financing arrangements. Sample sizes vary
between 250 and 1,500 companies per country and data are collected using either simple random
7
or randomly stratified sampling.
The sample includes formal enterprises of all sizes and
different ownership types across 26 industries in manufacturing, construction, services, and
transportation. Firms from different locations, such as the capital city, major cities, and small
towns, are included.
The use of firm-level survey data in cross-country work has become increasingly popular
in recent years and has several decisive advantages over the use of aggregate country-level
data. 16 First, existing papers using the same database show that firms’ responses to the survey
are closely related to measurable outcomes in terms of corruption, expropriation, property rights
protection, corporate financing, operation obstacles, tax evasion, investment, performance, and
growth (e.g., Johnson et al. (2000), Djankov et al. (2003), Acemoglu and Johnson (2005), Beck,
Demirguc-Kunt, and Maksimovic (2005), Beck, Demirguc-Kunt, and Levine (2006), Ayyagari,
Demirguc-Kunt, and Maksimovic (2008, 2010), Barth et al. (2009)). Second, the data set
provides unique and direct evidence on firm-level corporate tax evasion for a large sample of
firms across more than 100 countries around the world. Third, we are able to explore withincountry variation in tax evasion across firms of different types. Specifically, we are able to
compare firms of different sizes and in different locations, as well as firms from industries with
different financing needs. This enables us to apply a difference-in-difference approach to
alleviate endogeneity concerns and to explore specific mechanisms through which financial
outreach affects firm tax evasion.
Given the trade-off between data availability (e.g., availability of sampling weights and
panel dimension) and the cross-country scope of the sample coverage, we use several samples in
our analysis. Our first and broadest sample consists of 157 surveys across 102 countries,
collected over the period 2002 to 2010, with over 64,000 firm observations. This sample consists
of a broad cross-section of developing, emerging, and even several developed countries.
However, we have sampling weights for only a subset of these surveys. In robustness tests, we
therefore use a subsample of 21,500 firm observations with sampling weights in 38 surveys
across 34 countries during the period 2006 to 2010. This subsample controls for biases that might
8
arise from inconsistent sampling across countries or sampling that is based on any of our
explanatory variables, including firm size and location. To control for confounding time-variant
factors related to either global business cycles or changes within countries, we employ country ×
year fixed effects in some of our analyses. We also confirm all our findings with regressions that
only use data from the latest enterprise survey of each sample country.
Using these samples, however, only exploits cross-sectional variation, as even for
countries with several surveys, these are repeated cross-sections rather than panels. We therefore
use a panel of 85 surveys across 42 countries that allow us to exploit within-firm variation in tax
evasion and banking sector outreach. Unlike the larger cross-sectional samples, this is a
relatively small sample with around 3,800 firms. Surveys were undertaken between 2002 and
2010 and the sample includes countries from Europe & Central Asia, Latin America & the
Caribbean, the Middle East & North Africa, and Sub-Saharan Africa. This panel allows us to
exploit within-firm variation over time and thus control for omitted time-invariant firm-level
factors that might drive our findings in the cross-sectional regressions. This panel also allows us
to control for time-invariant firm-level measurement bias.
We construct the tax evasion variable using responses from the following question:
“Recognizing the difficulties many enterprises face in fully complying with taxes and regulations,
what percentage of total sales would you estimate the typical establishment in your area of
activity reports for tax purposes?” Using responses to this question, we construct two variables:
Tax Evasion Ratio, which is one minus the share of sales reported for tax purposes, and Tax
Evasion Dummy, which is one if a company’s tax evasion ratio is a nonzero positive number.
While the latter variable gauges the incidence of tax evasion, the former gauges the extent. The
tax evasion ratio ranges from an average of 68% in Gambia to less than 4% in Ireland, with an
average across countries of 21%. While in Brazil 83% of firms report tax evasion in their
industry, in Spain this figure is only 18.5% and the average across countries is 45%. Table I
reports the average values for these two indicators across the countries in our sample. The
variation in tax evasion is large both across and within countries, with a cross-country standard
9
deviation of 0.150 and a within-country standard deviation of 0.278. 17 The high within-country
variation suggests that it is important to focus on the firm-level rather than the country-level and
that there is important cross-firm and cross-industry variation to be exploited when gauging the
relationship between financial sector indicators and firm-level tax evasion. Similarly, the
variation in tax evasion is large both across and within country-industry cells. The cross-countryindustry standard deviation is 0.158 and the within-country-industry standard deviation is 0.272.
While many of the country-level numbers reported in Table I match anecdotal evidence, there
are also surprising findings, such as Germany, where 45% of firms state that they evade taxes
although the tax evasion ratio is rather low at 5.7%.
[Table I here]
The question on tax evasion is worded in this indirect way to elicit more honest answers.
Nevertheless, this wording might provide some measurement error as responses might reflect
perceived industry averages rather than own behavior. There are several reasons to believe that
this does not bias our results. First, as Johnson et al. (2000) point out, “managers presumably
most often respond based on their own experiences, and with caution we believe the responses
can be interpreted as indicating the firms’ own payments.” In fact, the large within-countryindustry variation also alleviates the concern that the firm responds to the question based on
industry average rather than own behavior. Second, tax evasion ratios are relatively stable over
time within a country. The correlation between tax evasion ratios from the Enterprise Surveys
over 2002 to 2010 and from the World Business Environment surveys in 1999/2000 is 64%.
Third, there is a high correlation between the ratio of informal activity to GDP and tax evasion.
Specifically, using data from Schneider and Ernste (2000) we find a correlation coefficient of
65%, significant at the 1% level. We also find a high correlation (>60%) between our tax evasion
measure and the tax evasion index developed by the World Competitiveness Yearbook. 18
We recognize that the measurement bias can go both directions, especially when the
interviewer is from the public sector or any financial institution with which the entrepreneur
might have a (potential) relationship. Fortunately, the World Bank is well aware of the potential
10
biases and hence organizes the interviews in a particular way to avoid these biases. For instance,
given the sensitive nature of the data, government officials are not directly involved in data
collection of these surveys nor are they given any raw data or any information that allows them
to identify the responses of individual firms. Critically, financial institutions are not involved in
the collection of such data – respondents are therefore unlikely to base their response on their
(potential) relationship with a financial institution. Rather, the surveys are conducted by the
World Bank in partnership with the local private sector such as independent chambers of
commerce or business associations that the local firms have confidence in. Further, questions on
the relationship between private and public sectors are at the end of the questionnaire once the
enumerator has established trust and confidence of the surveyed. In addition, the use of withinfirm variation over time reduces these biases, as we discuss below. A somewhat different
measurement concern is that we measure tax evasion only for existing formal enterprises, not
capturing informal enterprises; however, this will underestimates the variation in tax evasion
across countries (Johnson et al. (2000)).
We relate our measures of tax evasion to an array of financial sector indicators. We start
with a standard indicator of financial depth, Private Credit to GDP, which measures total
outstanding claims of financial institutions on the domestic nonfinancial private sector relative to
GDP (e.g., Bekaert, Harvey, and Lundblad (2005)). Previous research finds a positive and
significant relationship between financial sector depth and economic growth (Beck, Levine, and
Loayza (2000)). While Private Credit to GDP has been traditionally used as an indicator of
financial development, it does not properly measure the breadth of the financial system, that is,
the extent to which financial institutions cater to smaller and geographically more remote
customers. We therefore use a recently compiled data set on banking sector outreach – the
Financial Access Survey from the IMF. Specifically, we use Demographic Branch Penetration,
which is defined as the number of bank branches per 1,000,000 adults. We have data available
since 2003/4 and match the year of the firm-level survey to the year of branching data, with the
exception of firm-level surveys in 2002 for which we use 2003/4 branch penetration data. 19
11
While branch penetration is positively correlated with Private Credit to GDP, this correlation is
far from perfect. For example, both Bulgaria and Egypt have Private Credit to GDP ratios
around 47%, but Demographic Branch Penetration is 7.5 per 1,000,000 adults in Bulgaria while
it is 0.42 in Egypt. Beck, Demirguc-Kunt, and Martinez Peria (2007) show that higher branch
penetration is associated with a higher share of households and firms that use formal financial
services and with lower self-reported financing constraints of firms. 20
In addition to branch penetration, we use several indicators of the information framework
supporting the banking sector, as previous research shows the relevance of credit information
sharing especially for smaller firms (Brown, Jappelli, and Pagano (2009)).
Our principal
indicator is Depth of Credit Information Sharing, which ranges from zero to six and indicates
how much information on what share of the borrower population is collected and distributed, as
well as whether both financial and nonfinancial institutions are tapped for information.
Specifically, a value of one is added to the index when a country’s information agencies have
each of these characteristics: (1) both positive and negative credit information are distributed; (2)
data on both firms and individual borrowers are distributed; (3) data from retailers, trade
creditors, or utilities, as well as from financial institutions, are distributed; (4) more than two
years of historical data are distributed; (5) data are collected on all loans of value above 1% of
income per capita; and (6) laws provide for borrowers’ right to inspect their own data. 21 We use
the value of this variable for the same year as the respective firm-level survey. In the case of
firm-level surveys in 2002, we use the 2003 value, as this is the earliest year for which
information on credit information sharing was collected by the Doing Business team at the
World Bank.
We control for an array of firm characteristics that might be correlated with the decision
to underreport sales and that are defined in more detail in the Appendix. Specifically, we include
the size of the enterprise, as measured by the log of the number of employees, the log of firm age,
the location (capital city or small city/town, with medium-sized city as the omitted category), a
dummy variable if the firm is an exporter, and the share of foreign ownership. Finally, we
12
include a dummy variable indicating whether the firm’s financial statements are reviewed by an
external auditor. From theory and previous research, we expect size, age, exporter, and foreign
ownership to be negatively associated with tax evasion, while we expect firms located in smaller
towns to be more likely to evade taxes. 22 As indicated in Table II, 48% of the firms in our
sample are small firms (fewer than 20 employees), while 21% are large firms (more than 100
employees), with an average of 28 employees. On average, firms are 14 years old and the
average share of foreign ownership is 10%. Further, 20% of firms are exporting, 28% of firms
are in small cities and towns while 34% are in the capital city. Finally, 49% of the firms in our
sample have their financial statements audited.
We also include an array of country-level control variables. In addition to controlling for
financial depth, we include the indicator Bank Concentration, which is the Herfindahl index
using data from Bankscope. An extensive literature explores the relationship between market
structure and access to finance; 23 market structure, however, might also impact the degree of
credit information sharing among banks and their incentives for branch expansion. Controlling
for Private Credit to GDP and Bank Concentration increases our confidence that the measures of
branch penetration and credit information sharing do not capture other dimensions of financial
development. In addition, we control for GDP per capita, to discriminate between economic and
financial development. Our sample varies between Ethiopia with 123 U.S. dollars GDP per
capita and Ireland with a GDP per capita of more than 48,000 U.S. dollars (at the time of the
survey). As with all time-varying country-level variables, we use the value for the same year as
the respective firm-level survey.
We also include several proxies for alternative explanations of tax evasion. First, we
include Tax Rate, which is measured as the total tax payments and contributions of a typical
commercial enterprise relative to its profits (Djankov et al. (2010)). On the one hand, a higher
tax rate increases the benefits of evasion. On the other hand, a higher tax burden can also lead to
better public good provision by the government, which increases the opportunity costs of tax
evasion. We include an indicator that captures the time it takes to prepare, file, and pay corporate
13
income tax, value added or sales tax, and labor taxes, including payroll taxes and social
contributions. We also include an indicator for the number of tax payments a typical company
has to make per year and the ease with which they can be done. These two indicators proxy for
the tax administration burden firms face and should be positively associated with tax evasion.
We also include a dummy for countries that have a VAT system in place. A VAT system allows
enterprises to offset tax payment on inputs against tax payments on sales and should thus reduce
benefits from tax evasion. A VAT system is in place for 89% of countries. 24 Second, as pointed
out in the literature (e.g., Friedman et al. (2000), Johnson et al. (2000)), tax evasion might be
driven by bureaucracy, predatory behavior by government officials such as bribery seeking,
weak legal institutions, and deficient public services. Therefore, in our baseline regressions, we
include the country-level indicators Rule of Law, Control of Corruption, and Government
Effectiveness, all from the Kaufmann, Kraay, and Mastruzzi (2010) Governance Matters database
and its updates. We also control for the country-level crime rate as firms might hide some of
their profits to escape extortion by criminal gangs (Johnson et al. (2000)). Detailed definitions of
these variables can be found in the Appendix.
Table II presents descriptive statistics of all variables, while in the Internet Appendix we
report the correlations between the different variables. 25 The correlations indicate that firms
located in smaller towns, smaller firms, and younger firms evade a higher share of taxes, while
foreign-owned firms and exporting firms evade taxes to a lesser degree. However, there are also
many significant correlations between firm characteristics. Smaller firms are more likely to be
located in smaller towns and are less likely to be an exporter, and are younger. The countrylevel correlations show that tax evasion by firms is more prominent in countries with lower
branch penetration and less efficient credit information sharing. However, tax evasion is also
significantly associated with corruption, taxation, government effectiveness, and economic and
financial development, thereby underscoring the need for multivariate analysis.
[Table II]
B. Methodology
14
To assess the relationship between tax evasion and banking sector outreach, we run the
following regression:
𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝐹𝐹𝑖𝑖 + 𝛽𝛽𝐶𝐶𝑖𝑖 + 𝛾𝛾𝐵𝐵𝑗𝑗 + 𝜄𝜄𝑘𝑘 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖,
(1)
where T is the tax evasion ratio or dummy as reported by firm j in country i and industry k, F is a
vector of financial sector indicators, including indicators of credit information sharing and
branch penetration, C is an array of country-level control variables, B is a vector of firm-level
control variables, as discussed above, ι is a vector of 26 industry dummies, and ε is the error
term. We also include year dummies for the year in which the survey was conducted to control
for any global trends and for differences within countries with several surveys. We use a Tobit
model for the regression of tax evasion ratio, as this variable is bounded between zero and one,
and a Probit model for the regressions of tax evasion dummy. We report marginal effects rather
than coefficient estimates to gauge the statistical as well as economic significance of our
regression results. Further, in the spirit of Petersen (2009), we report p-values based on clustered
standard errors, that is, allowing for correlation between error terms within-country, but not
across countries. A negative and significant α would indicate that deeper financial systems,
higher branch penetration, and a more effective and inclusive information framework are
associated with a lower incidence of tax evasion and a lower tax evasion ratio. As these
estimations are subject to endogeneity and omitted variable biases, we also present instrumental
variables (IV) and firm fixed effects regressions, discussed in more depth below.
The variation across firms of different sizes, locations, and sectors allows us to test for a
differential impact of financial sector development on tax evasion. Specifically, the hypotheses
formulated above predict the impact of financial sector development to be stronger for smaller
firms and for firms in more remote locations. 26 We test for such differential impact by
augmenting equation (1) with interaction terms in the following regression models:
and
𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝐹𝐹𝑖𝑖 + 𝛽𝛽𝐶𝐶𝑖𝑖 + 𝛾𝛾𝐵𝐵𝑗𝑗 + 𝛿𝛿𝐹𝐹𝑖𝑖 ∗ 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑗𝑗 + λ𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑗𝑗 + 𝜄𝜄𝑘𝑘 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖,
(2)
𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝐹𝐹𝑖𝑖 + 𝛽𝛽𝐶𝐶𝑖𝑖 + 𝛾𝛾𝐵𝐵𝑗𝑗 + 𝛿𝛿𝐹𝐹𝑖𝑖 ∗ 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 + λ𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝑗𝑗 + 𝜄𝜄𝑘𝑘 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖,
(3)
15
where Size is a vector of dummies for small and large firms (with medium-sized firms being the
benchmark category) and Location is a vector of dummies for firms in the capital city and small
cities (with firms in medium-sized cities being the benchmark category). A small city is defined
as having less than 250,000 inhabitants. The linear term of Size and Location are also included in
the model. Theory would suggest a negative coefficient on the interaction of financial sector
depth and outreach with Small Firm and Small City, while we expect positive coefficients on the
interaction of financial sector depth and outreach with Large Firm and Capital City.
Beyond size and location influencing firms’ benefits from formality in countries with
more effective credit information sharing and higher branch penetration, there might also be
industry variation and time-variant country-level variation in these benefits. A large literature
exploits industry variation in characteristics such as dependence on external financing and
growth opportunities as identification conditions to assess the impact of financial and
institutional development on firm growth.
Such an identification strategy relies on the
assumption that such industry features are constant across countries and uses actual data on
external financing and growth from industries in the U.S. as benchmarks under the assumption
that they reflect demand-side factors. 27 Unlike firm characteristics that might be endogeneous to
the perceived benefits of tax evasion, these industry characteristics are exogenous. We focus on
two industry characteristics constructed based on these assumptions from existing literature.
First, Dependence on External Finance is the fraction of capital expenditures not financed with
internal funds (Rajan and Zingales (1998)). We use data from the Capital IQ database, which
contains a large number of both public and private firms in the U.S., to construct this indicator.
Second, we use two measures of growth opportunities. The first measure follows Fisman and
Love (2007) and takes U.S. industrial value-added growth over the period 1990 to 1999 as gauge
for growth opportunities outside the U.S. over our 2002 to 2010 sample period. A second
measure of growth opportunities is at the country level. Following Bekaert et al. (2007), we
compute the exogenous growth opportunities of country i in year t as the PE ratios computed on
global data on listed companies, averaged across 35 sectors weighed by annual country-specific
16
industry weights based on lagged market capitalization. As this measure might be driven by
differences in persistent discount rates, we follow Bekaert et al. (2007) and remove a 60-month
moving average from this measure. Detailed definitions of the variables can be found in the
Appendix.
To test for a differential impact of banking sector outreach on firms in different industries,
we use the following specification:
𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝐹𝐹𝑖𝑖 + 𝛽𝛽𝐶𝐶𝑖𝑖 + 𝛾𝛾𝐵𝐵𝑗𝑗 + 𝛿𝛿𝐹𝐹𝑖𝑖 ∗ 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑘𝑘 + 𝜄𝜄𝑘𝑘 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖,
(4)
where Industry is an industry characteristic, either dependence on external finance or growth
opportunities. 28 Since we control for industry dummies and include the levels of the respective
financial sector indicators, the δ coefficients capture the differential effect of credit information
sharing and branch penetration on firms in industries with different financing needs and growth
opportunities.
While we report Tobit regressions to assess the differential impact of size, location, and
industry characteristics on the relationship between branch penetration, credit information
sharing, and tax evasion, we confirm all our findings with OLS regressions given the difficulty
of interpreting the marginal effects of interaction terms in nonlinear models (Ai and Norton
(2003)). We also present IV regressions for models (2) to (4) where we instrument for credit
information sharing, branch penetration, and their interaction with firm, industry, and country
characteristics.
In a final set of regressions, we use a smaller panel sample of firms and countries to test
the relationship between credit information sharing, branch penetration, and tax evasion over
time:
𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝛼𝛼𝐹𝐹𝑖𝑖 , 𝑡𝑡 + 𝛽𝛽𝐶𝐶𝑖𝑖, 𝑡𝑡 + 𝛾𝛾𝐵𝐵𝑗𝑗, 𝑡𝑡 + 𝛿𝛿𝛿𝛿𝑗𝑗 + 𝜑𝜑𝐹𝐹𝑖𝑖, 𝑡𝑡 ∗ 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝑗𝑗 + λ𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝑗𝑗 + 𝜀𝜀𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖,
(5)
where Xj are firm fixed effects and t is the year of the survey. Year dummies are also included.
Here, we only include the time-varying firm variables among the vector B of firm-level
characteristics. We focus on regressions where we interact credit information sharing and branch
penetration with Firm, which is an array of firm size dummies, firm location dummies or
17
industry characteristics. Unlike the preceding regressions, we use OLS to estimate specification
(5), given that Tobit panel data models with fixed effects yield biased estimates (see Greene
(2004)). 29 We also run IV regressions, where we instrument for credit information sharing,
branch penetration, and their interaction with firm and industry characteristics.
II. Empirical Results
Combining firm-, industry- and country-level variations, this section tests whether better
credit information sharing and higher branch penetration are associated with lower tax evasion.
We first explore the effect of cross-country variation in credit information sharing and branch
penetration, before combining it with firm- and industry-level variations.
We also use
instrumental variable analysis and firm-level fixed effects regressions for a sub-sample to control
more rigorously for simultaneity and endogeneity biases and gauge the robustness of our
findings for several subsamples. Throughout the discussion, we report both statistical and
economic significance of our results.
A. Baseline Results
The results in Table III show a statistically and economically significant relationship
between the depth of credit information / branch penetration and the incidence and extent of tax
evasion across a sample of 157 surveys and 102 countries. We report both Probit (Panel A) and
Tobit regressions (Panel B) that include unreported industry and year dummies and the standard
errors of the coefficients clustered at the country level. We present three Probit and three Tobit
regressions, where we start with a simple model with only the two indicators of depth of credit
information and branch penetration, GDP per capita, and firm-level control variables. We then
add country-level variables related to taxation before finally adding other country-level variables
that might explain cross-country variation in the incidence and extent of tax evasion.
[Table III here]
As can be seen from Table III, the depth of credit information and branch penetration are
18
associated with a lower incidence and extent of tax evasion. Both Depth of Credit Information
Sharing and Demographic Branch Penetration enter negatively and significantly in the Probit
and Tobit regressions. The effect is also economically significant. The results in columns (1) and
(4) suggest that a one-standard-deviation increase in Depth of Credit Information Sharing is
associated with a 16.6% drop in the likelihood of corporate tax evasion and a 12.6% drop in the
tax evasion ratio, while a one-standard-deviation increase in Demographic Branch Penetration is
associated with a 12.3% reduction in the incidence of tax evasion and a 9% reduction of the tax
evasion ratio. 30 Given the sample mean of the incidence of tax evasion (46.3%) and the extent of
tax evasion (21.3%), these effects are quite substantial. We also note that neither statistical nor
economic significance is affected as we include country-level control variables.
Turning to the control variables, we find that GDP per capita enters consistently with a
negative sign, though not always significantly, suggesting that higher levels of economic
development are associated with a lower incidence and extent of tax evasion. Several of the firmlevel variables enter significantly in the regressions, including firm size, location, ownership, and
the dummy indicating whether firms’ financial statements are audited. We also find that several
dimensions of the tax system are significantly and positively associated with tax evasion,
including the total tax rate, the time to prepare and pay taxes, and the total number of taxes. On
the other hand, firms in countries with a VAT system report lower tax evasion, though this
relationship is not always significant. Concerning alternative country-level explanations of tax
evasion, we find that deeper financial sector depth, as proxied by Private Credit to GDP, is
associated with a lower incidence and extent of tax evasion. A higher bank concentration is
positively and significantly associated with the extent but not the incidence of tax evasion. Better
rule of law, tighter control of corruption, more effective government, and lower crime all reduce
tax evasion, though not always significantly.
B. Instrumental Variable Analysis and Sample Weights
Endogeneity is often a concern in cross-country studies. In our study, it is conceivable
19
that a high degree of tax evasion could generate calls for a higher degree of information sharing
and branch expansion. If this kind of feedback from the corporate sector to policymaking were in
force, we should observe a positive relation between depth of credit information / branch
penetration and tax evasion. However, we find a strong and negative relation between
information sharing / branch penetration and tax evasion, confirming that reverse causality might
not be a first-order concern in this study. Nevertheless, since our baseline regressions link
country variation in banking sector outreach to firm-level tax evasion, omitted variable bias
could still be of concern. We address this concern with instrumental variable regressions and
report both first- and second-stage results, both for the larger sample as well as for the sample
with sampling weights.
To address the concern of possible reverse causality between financial outreach and tax
evasion, we conduct a thorough search of related documents, reports, announcements, papers,
and books about the potential driving factors of the establishment of credit registries and the
expansion of branch networks. We do not find any quotes about tax evasion as a reason for these
activities. Instead, we find the major reasons for setting up credit registries and the expansion of
branch networks are to improve the credit assessment and risk management of financial
institutions, strengthen bank supervision and regulation, lower the cost of information collection,
enhance financial stability, and facilitate access to credit markets. Therefore, in this study we
use measures of the banking regulatory and supervisory structure as instrumental variables for
depth of credit information sharing and branch penetration. The data are obtained from the Bank
Regulation and Supervision Database compiled by Barth, Caprio and Levine (2006, 2008).
Detailed definitions of the variables can be found in the Appendix. Intuitively, bank regulation
and supervision should have a direct impact on banks’ risk management incentives and outreach
decisions but not affect corporate tax avoidance directly. To reduce the risk that these regulatory
and supervisory structures are due to policy reforms taken at the same time as changes in credit
information sharing systems, we focus on variables that are persistent or that do not directly
reflect policy actions.
20
We use several variables gauging the structure, strength, and independence of
supervisory authorities. Specifically, we include the average tenure of bank supervisors as a
proxy both for experience and independence, the log of the number of bank supervisors, an index
of supervisory independence from both banks and politicians, and an index of supervisory power
vis-a-vis banks in good and bad times. More experienced and independent bank supervisors are
more likely to adopt state-of-the-art bank regulations that help improve banking system stability
(Houston, Lin and Ma (2012)). This might also be the case for countries with more complicated
regulatory systems (e.g., the presence of multiple bank regulators). More powerful supervisors
are more capable to push for reforms and adopt new regulatory frameworks. We also use the
historical nonperforming loan ratio (i.e., the previous five-year average ratio of nonperforming
loans to total loans) as an additional instrumental variable. A high nonperforming loan ratio
might increase banks’ and the regulator’s incentives to adopt regulation and mechanisms that
enhance banks’ risk management and credit assessment practices (e.g., credit registries).
Furthermore, following Demirguc-Kunt and Detragiache (2002), we also use an indicator of
“policy contagion” by including the share of countries in each region with a credit registry. As
regulators or policymakers learn more about the workings of a regulation from those countries
implementing the regulation, they might modify their regulations after observing regulatory
changes in other countries in the same field. In addition, we include the share of foreign and
private bank ownership as additional instrumental variables. Unlike state-owned banks, privately
owned (especially foreign) banks are typically more efficient in their outreach, have stronger
incentives to manage credit risk, and rely more on hard information and thus credit registries.
The first-stage results presented in the Internet Appendix show significant relationships
between the instrumental variables (IV) and our indicators of credit information sharing and
branch penetration. The findings are largely consistent with our expectations. Specifically, we
find that countries with supervisors that have longer tenure and enjoy more independence and
power, higher ratios of nonperforming loans, more private and foreign bank ownership, and a
larger share of countries in the same region with credit registries have higher Depth of Credit
21
Information Sharing. Countries with supervisors with longer tenure and higher independence, a
larger private and foreign bank market share, and a bigger share of neighboring countries in the
region with credit registries have higher Demographic Branch Penetration. While not all
variables enter significantly in all four first-stage regressions, they enter jointly significantly at
the 1% level in all eight regressions, with F-values well above 10, a threshold often used to assess
the relevance of instruments (Staiger and Stock (1997)). In robustness tests, we include other
variables related to supervisory structure and banks’ accounting transparency and obtain similar
results for both first- and second-stage regressions.
The second-stage regression results reported in Table IV Panel A show the robustness of
our findings to controlling for endogeneity and simultaneity biases. Henceforth, we report only
Tobit regressions with the tax evasion ratio, though our results are confirmed when using Probit
regressions with the incidence of tax evasion. We find very strong and consistent evidence that
more effective credit information sharing and higher branch penetration are associated with
lower degrees of tax evasion. The IV coefficients are also of similar size as the OLS coefficients.
We also conduct the Hansen overidentifying J-test to assess whether the instrumental variables are
associated with the tax evasion ratio beyond their effects through depth of credit information,
branch penetration, or other explanatory variables. The Hansen’s J-test suggests that we cannot
reject the null hypothesis that the instruments are valid in all model specifications. 31
[Table IV here]
The results in Table IV Panel B show that our results are robust to controlling for sampling
weights using a smaller sample of countries for which such weights are available, while at the
same time using instrumental variables.32 The sampling weight is the inverse of the probability
of selection of the firms surveyed. The interpretation of the sampling weight is that it
characterizes the number of firms in the population that are represented by this particular sample
firm. Only Enterprise Surveys from 2005/6 onwards include sampling weights, in total, 38
surveys for 34 countries with 21,541 observations available in our IV regressions. The empirical
results presented in Panel B of Table IV show that our main findings are confirmed, with the
22
coefficients on our main variables of interest entering with similar statistical and economic
significance as in Panel A of Table IV. The first-stage results in the Internet Appendix again
confirm the relevance of our instruments.
We use an additional test to control for possible differences in sampling methodologies
across regions. Specifically, we test the robustness of our results by dropping one region at a
time - 1) East Asia & the Pacific, 2) Europe & Central Asia, 3) Latin America & the Caribbean,
4) the Middle East & North Africa, 5) South Asia, and 6) Sub-Saharan Africa – from the
regressions in Table III. The empirical results are highly robust.
While the IV regressions control for endogeneity and omitted variable biases, we test
directly for reverse causation by relating changes in credit information sharing and branch
expansion between the two surveys to the average level of tax evasion in each country, reported
in the Internet Appendix. Tax evasion enters insignificantly in all regressions, suggesting that
policy changes to credit registries and bureaus and branch expansion by banks are not related to
the ex-ante tax evasion environment. In addition, we estimate the correlation between changes in
credit information and changes in general institutional environment measures (e.g., government
effectiveness, control of corruption, quality of regulation, rule of law, property right index) and
find very low correlations (all the correlation coefficients are below 0.1 and are not statistically
significant). This helps alleviate the concern that changes in information sharing may be part of a
larger institutional reform.
In summary, our results are robust to the use of instrumental variables and sampling
weights and there is no evidence of reverse causation or simultaneity bias. Nevertheless, we
regard the IV regressions as one of several different techniques to control for endogeneity and
omitted variable bias and present other tests below.
C. Further Robustness Tests
The Internet Appendix reports an array of robustness tests with different sample cuts and
additional control variables. First, we drop 21 countries from our sample that either have Islamic
23
banks (as defined in Bankscope) or have adopted Islamic law, as defined by the CIA Factbook. 33
Unlike conventional banking, Islamic banking relies more on risk-sharing arrangements and
close relationships between the lender and borrower and thus less on formal credit informationsharing arrangements. Our findings are confirmed for this subsample of countries. Second, we
control for the proportion of informal financing in total financing of working capital and
investment finance. In countries with underdeveloped financial systems, informal financing
sources substitute for formal bank credit, increasing the likelihood that the relationship between
tax evasion and financial sector outreach is spurious. Our main findings are not affected.
Third, the literature points out that tax evasion might also be caused by the interplay
between the state and organized crime (e.g., the mafia) in the provision of public goods (e.g.,
Grossman (1995), Alexeev, Janeba and Osborne (2004)). In countries with inefficient
governments, the mafia might partially substitute for the provision of public goods. As a
consequence, firms have incentives to evade taxes, instead paying these resources to the mafia
for the provision of public goods (e.g., contract enforcement). If the above effect is more
prevalent in those countries with a lower degree of branch penetration and depth of credit
information, the relationship between tax evasion and depth of credit information / branch
penetration might be a spurious one. Therefore, we limit our sample to firms that do not pay for
mafia protection and whose tax evasion is not driven by the channels we discussed above. 34 We
restrict our sample to countries with above-median rule of law, as firms in these countries are
less likely to rely on the mafia for provision of public goods. We also confine our sample to
countries with high government effectiveness as firms are less likely to rely on the mafia for
public good provision in these countries. In all three cases, we confirm our findings with similar
statistical and economic significance.
Fourth, we test for countervailing effects by gauging whether our results hold for two
subgroups of firms. In general, the more wealth a firm hides in tax evasion, the less collateral it
can offer for securing a loan and the lower is the likelihood of getting access to credit with
reasonable terms and conditions. However, in the case of private firms, the share of profits
24
hidden from the tax authorities remains with the business owner, increasing the value of cash
collateral the owner can provide to banks to secure future loans. Tax evasion in the case of public
companies, on the other hand, does not accrue to owners, but rather to management, increasing
agency problems within firms. Therefore, we limit our sample to publicly traded firms and find
that the results remain significant both statistically and economically. Moreover, as documented
in some recent reports (e.g., Gravelle (2010)), large multinational firms and wealthy business
owners might have various ways to avoid taxes such as shifting profits into low-tax foreign
subsidiaries, shifting debt to high-tax jurisdictions, and setting up secret bank accounts in tax
haven countries. To ensure that our results are not driven by these alternative tax evasion
channels, we focus on domestic small firms (e.g., firms with less than 20 employees and with no
foreign subsidiaries), which are less likely to engage in sophisticated international tax avoidance
activities. We continue to find that our main findings on financial outreach and tax evasion are
statistically and economically significant.
Fifth, we address the concern that changes in credit information sharing and branch
networks might occur simultaneously with other economic policy reforms in the financial system,
legal environment, and monitoring mechanisms to reduce tax evasion, often fostered by
comprehensive reform packages with IMF or World Bank support. We therefore drop any
country that received IMF financial aid or a World Bank funded project in the areas of the
financial sector, fiscal policy, public administration, or legal system one year prior to and also
during the survey years. In the case of several surveys in a country, we also require that there
was no IMF or World Bank assistance between the survey years. We further control for a capital
account liberalization index and for an equity market liberalization index to alleviate the concern
that changes in tax evasion are driven by other types of capital market reforms. In both cases, the
results are consistent with our expectation. Finally, using the panel data set on top marginal tax
rates compiled by the Fraser Institute, we confirm our findings for a subsample of 61 countries
without changes in tax rates one year prior to, and also during, the survey years. Overall, these
tests further strengthen our main findings and mitigate the concern that the empirical findings are
25
driven by other capital market or tax reforms.
Further regressions in the Internet Appendix confirm the robustness of our findings to
using alternative measures of the information-sharing framework and to controlling for an array
of additional institutional indicators. Specifically, dummy variables for the existence and
coverage ratios of both private and public credit registries are associated with lower tax evasion
ratios, with the economic size of the effects being similar. Our findings are also robust to
controlling for other dimensions of a country’s institutional framework, including summary
indicators of voice and accountability, political stability, and regulatory quality, as well as the
number of registration steps for new businesses, creditor rights, a property right index, and an
index of the risk of expropriation by the government. 35 While many of these indicators enter
significantly and with the expected signs, depth of information sharing and geographic branch
penetration (an alternative measure of banking penetration defined as the number of bank
branches per 10,000 sq km) continue to enter negatively and significantly in all regressions.
The Internet Appendix provides an additional robustness test related to possible
measurement error within country-industry cells. The data on tax evasion are based on a survey
question related to industry practices. While this question is expected to capture the actual
behavior of the surveyed firm, there might be cross-firm variation on whether a respondent refers
to own or competitors’ behavior. To control for this concern, we re-run our main regressions at
the industry level, averaging firm-level variables within each survey-industry cell. This provides
us with 1,490 observations across 102 countries and 26 industries, with on average 43 firms per
cell. The analysis replicates the Table III Tobit regressions for the industry-level sample and
confirms our main findings.
We undertake several additional robustness tests. First, we limit the sample to 33
countries with at least two surveys and changes in the country-level variables, thus focusing on
the within-country variation in the relationship between tax evasion and financial sector outreach.
Including country fixed effects allows us to control for other time-invariant country-level omitted
variables. Second, we test whether our results are driven by one specific country and replicate
26
the Table III results omitting each country, one at a time. Since the relationship between credit
information sharing, branch penetration, and tax evasion might vary with income level, we also
drop all 15 high-income countries 36 from our sample to re-run our regressions. Finally, we limit
our sample to the latest survey for each country, which reduces our sample to 18,500 firms,
confirming our results.
Up to now we have related firm-level responses to country-level variation in credit
information sharing and branch penetration. However, different firms might react differently to
the incentives and opportunities provided by better credit information sharing and banking sector
outreach. We explore this possibility in the following; testing for such differential impact also
allows us to more rigorously address the issue of omitted variables and causality.
D. Exploiting Firm Location Heterogeneity
The hypotheses formulated in the introduction suggest a differential relationship of
information sharing and branch penetration with firms’ decision to evade taxation across firms in
different locations. Specifically, firms in more remote areas are conjectured to respond more
strongly to incentives and opportunities provided by more effective information sharing and
branch penetration than firms in capital cities. We examine the firm location conjecture and
present the empirical results in Table V.
[Table V here]
The results in Table V confirm our conjecture and show a significant impact of firm
location on the relationship between information sharing, branch penetration, and firms’ decision
to evade taxes. Here we add interaction terms of Depth of Credit Information Sharing and
Demographic Branch Penetration with dummy variables that indicate whether a firm is located
in the capital city or a small town, using firms in mid-sized towns as the omitted category. While
we find a more muted relationship between information sharing, branch penetration, and tax
evasion for firms in the capital city, the relationship is even stronger for firms in small towns.
The differences in the relationship across firms of different locations are also economically
27
significant. A one-standard-deviation increase in Depth of Credit Information Sharing decreases
tax evasion by 8.1% for firms in the capital city, but by about 18.3% for firms in small towns
(column (1)). Similarly, a one-standard-deviation increase in Demographic Branch Penetration
decreases tax evasion by 7.9% for firms in the capital city, but by about 15.5% for firms in small
towns (column (2)). We also control for the interaction of Private Credit to GDP with the
location dummies. Compared to the location interaction terms with credit information depth and
branch penetration, however, the interaction of firm location with financial depth is small in size,
suggesting only a small differential impact of financial depth on firms in different locations.
The Table V regressions also show that our findings are robust to using IV regressions
and country fixed effects. In columns (3) and (4), we instrument for credit information depth and
branch penetration and their interaction with the small and capital city dummies with the same
variables as in Table IV and confirm our findings. In columns (5) and (6), we drop all countrylevel variables, including our financial sector indicators, and replace them with country-year
dummies. This allows us to control even more rigorously for confounding country factors. The
empirical results are highly robust. The interaction terms of firm location with branch penetration
and credit information depth enter significantly. We confirm our findings while controlling for
the interactions of Private Credit to GDP with firm location dummies.
E. Exploiting Firm Size Heterogeneity
Similarly, we also expect a differential relationship of credit information depth and
branch penetration with firms’ decision to evade taxation across firms of different sizes.
Presumably, smaller firms tend to respond more strongly to incentives and opportunities
provided by more effective information sharing and more expansive branch networks than bigger
firms. We test this hypothesis in Table VI.
[Table VI here]
The results in Table VI show a significant impact of firm size on the relationship between
depth of credit information, branch penetration, and tax evasion. A one-standard-deviation
28
increase in Depth of Credit Information Sharing decreases tax evasion by 6.6% for large firms,
but by about 16.9% for small firms (column (1)). Similarly, a one-standard-deviation increase in
Demographic Branch Penetration lowers tax evasion by 9.1% for large firms, but by about
14.2% for small firms (column (2)). The results in Table VI also reveal that our findings are
robust to (i) using IV regressions (columns (3) and (4)) and (ii) controlling for country-year fixed
effects (columns (5) and (6)). We also control for the interactions of Private Credit to GDP with
firm size dummies and find that our key results in Table VI are highly robust.
F. Exploiting Industry Heterogeneity
As discussed above, we expect a stronger link between depth of credit information
/branch penetration and tax evasion for firms in industries more dependent on external finance
and with better growth prospects. We also expect a stronger relationship between financial sector
outreach and tax evasion in countries with exogenously higher growth opportunities. The results
in Table VII indeed show significant industry heterogeneity in the relationship of credit
information depth, branch penetration, and tax evasion. Here, we interact an industry
characteristic (external finance dependence or growth opportunities) with our financial sector
indicators. The regression in column (1) suggests that the effect of Demographic Branch
Penetration and of Depth of Credit Information Sharing on reducing tax evasion increases in
firms’ dependence on external finance. Specifically, a one-standard-deviation increase in Depth
of Credit Information decreases tax evasion for firms in the most financially dependent industry
by 16.1% more than for firms in the least financially dependent industry. A one-standarddeviation increase in Demographic Branch Penetration decreases tax evasion for firms in the
most financially dependent industry by 16.5% more than for firms in the least financially
dependent industry (column (1)). Similarly, the column (2) and (3) regressions show that the
relationship of credit information depth, branch penetration, and tax evasion becomes much
stronger in sectors (GO1) and countries (GO2) with higher growth opportunities. Taken together,
the results suggest that financial sector outreach increases incentives for firms that are more
29
dependent on external finance and have higher growth opportunities to reduce tax evasion. The
empirical results in Table VII also show that our findings are robust to (i) controlling for
endogeneity and simultaneity biases by using instrumental variables (columns (4) to (6)) and (ii)
controlling for country-year-fixed effects (columns (7) to (9)).
[Table VII here]
G. Exploiting Time-series Variation
In this section, we exploit time-series variation in credit information depth and branch
penetration across a sample of 3,800 firms (7,671 firm-year observations) in 42 economies for
which we have unbalanced firm-level panels. As discussed above, this sample consists of
countries in Eastern Europe and Central Asia, Latin America, and Africa, with surveys between
2002 and 2010. This sample includes countries both with and without changes in credit
information depth, while almost all of them saw changes in their branch penetration over the
sample period. In the Internet Appendix we present the 15 countries that have seen changes in
credit information depth, with the years when these changes were undertaken. On average, the
magnitude of the changes in credit information depth is 1.59 (out of a maximum value of six).
Most notably are Georgia and Estonia, which went from having no credit registry to having an
almost full-fledged credit bureau (depth of credit information =5). Similarly, we also find
evidence that change in financial outreach is associated with firm tax evasion. For instance,
Kyrgyz Republic’s Demographic Branch Penetration increased from 0.461 to 0.533, and its
Depth of Credit Information Sharing increased from zero to two between 2002 and 2005 (the
two years when its two surveys were conducted). During the same period, Kyrgyz Republic’s
average tax evasion ratio declined from 25% in 2002 to 15% in 2005 and tax evasion incidence
declined from 58% in 2002 to 43% in 2005.
As we include firm fixed effects we drop firm characteristics except for the log of the
number of employees, firm age, and auditing status. Since panel Tobit estimates with fixed
effects tend to be biased (Greene (2004)), we use OLS regressions for our fixed effect panel
regressions.
30
The results in Table VIII show a negative relationship between financial sector outreach
and tax evasion for the sample of countries with a panel data set of firm surveys. We first report
OLS regressions without fixed effects in columns (1) to (4). We find that both credit information
depth and branch penetration are negatively and significantly associated with tax evasion. We
also find that the relationship between credit information depth and tax evasion is stronger for
firms in small cities, smaller firms, firms in industries that rely more on external finance (column
(1)), and firms in industries and countries with higher growth opportunities (columns (2) and (3)).
Finally, we confirm our findings for a smaller sample of 15 countries that experienced changes in
all country-level variables between the two surveys (column (4)).
The results in columns (5) to (8) show the robustness of our results to controlling for
time-invariant firm characteristics by including firm fixed effects. We thus estimate the withinfirm effect of changes in credit information sharing and branch penetration on the extent to
which firms evade taxes as well as the differential effects on firms in different locations, firms of
different sizes, firms across industries with different needs for external finance (column (5)), and
firms across industries and countries with different growth perspectives (columns (6) and (7)).
Overall, our previous findings are confirmed, in terms of both statistical and economic
significance. We confirm these findings for a smaller set of countries that experienced changes in
all macroeconomic variables between the two surveys in column (8), though some of the
variables enter less significantly. Finally, we confirm our findings controlling for endogeneity
and simultaneity biases by using instrumental variables for credit information depth and its
interaction with firm and industry characteristics (columns (9) – (12)).
[Tables VIII and IX here]
The results in Table IX confirm our finding that an increase in branch penetration is
positively associated with a reduction in tax evasion and more so for geographically more remote
firms, smaller firms, firms in industries with a higher need for external finance and higher
growth opportunities, and firms in countries with higher growth opportunities. As in Table VIII,
we first report simple OLS, before including firm fixed effects and instrumental variables. We
31
find that firms reduce tax evasion as Demographic Branch Penetration increases, and more so if
they are in smaller cities, have smaller size, are in industries with a higher need for external
finance (column (1)) or higher growth opportunities (column (2)), and are in countries with
higher growth possibilities (column (3)). The results are again confirmed when we focus on the
smaller sample of countries with changes in all country-level variables (column (4)). We confirm
our findings controlling for firm fixed effects (columns (5) to (8)) and with instrumental
variables (columns (9) to (12)). In summary, even when limiting our sample to countries for
which we can observe changes in tax evasion for the same firm over time and that experienced
changes in credit information sharing and branch penetration, we confirm our findings, including
for the differential effects across firms of different sizes, location, financing needs, and growth
opportunities.
H. Additional Results and Broader Implications
This section discusses several additional results that support our hypotheses. To save
space, these results are presented in the Internet Appendix.
The results show that firms in countries with more effective credit registries and a higher
level of branch penetration are less likely to report the severity of collateral as a constraint for
their operations and growth and are more likely to have audited financial statements. As
discussed in the introduction, the presence of collateral should be less of a constraint for access
to finance in economies with better credit information sharing and higher branch density,
because the information gap between creditors and borrowers is smaller and because creditors
can monitor firms more effectively. We therefore run ordered Probit regressions using
categorical responses to the survey question “Is access to finance (e.g., collateral) a problem for
the operation and growth of your business?” as the dependent variable. 37 The results suggest that
firms in countries with more effective systems of credit information sharing and higher branch
penetration are less likely to complain about collateral as a constraint for their operations and
growth. The hypotheses discussed in the introduction also suggest that the benefits of audited
32
financial statements or the costs of “cooking the books” might be higher in countries with more
effective credit registries and higher branch penetration. We therefore run a Probit model with a
dummy variable for whether firms have audited financial statements. The results show that firms
in countries with more effective systems of credit information sharing and higher branch
penetration are more likely to have audited financial statements.
Results reported in the Internet Appendix further suggest that tax evasion is negatively
associated with expected firm growth. While so far we have shown a robust relationship between
financial sector outreach and tax evasion, it is not clear whether there are performance
consequences of the firm’s decision. While a rigorous exploration of the link between tax
evasion and firm performance is beyond the scope of this paper, we provide some tentative
evidence on the negative consequences of tax evasion for firm performance by relating tax
evasion to expected sales growth. However, expected growth data are not available in the
Enterprise Surveys. These data are only available from a predecessor survey, the World Business
Environment Survey, undertaken in 1999/2000 across 80 countries. Using data for over 6,000
firms from this survey, we relate the expected sales growth rate over the next three years to a
categorical tax evasion variable.
38
We report simple correlations and then add firm
characteristics, including government and foreign ownership dummies, exporter and government
subsidy dummies, and the number of competitors, as well as industry fixed effects. Finally, we
add country fixed effects. Across all three specifications, we find a negative and significant
relationship between tax evasion and expected sales growth. This relationship is not only
statistically but also economically significant: an increase in ten percentage points in the tax
evasion ratio is associated with one percentage point lower expected growth over the following
three years, where the average expected growth rate is 21.7%. Therefore, tax evasion not only is
a corporate decision weighing benefits and costs, but also carries growth costs for the enterprise.
This result also reaffirms that our findings are unlikely to be driven by a correlation between
provision of public services by the mafia and financial sector infrastructure, as public services
provided by the mafia would not imply a negative impact on expected growth.
33
III. Conclusions
This paper explores the association of credit information sharing and bank branch
penetration with the incidence and extent of tax evasion across countries and firms. We find
strong evidence that firms in countries with deeper and more effective systems of credit
information sharing and higher bank branch penetration tend to hide a smaller share of their sales
and are less likely to evade taxes. This effect is particularly strong for firms in small cities and
for firms of smaller size. Furthermore, we find variation in the relationship between financial
sector outreach and tax evasion across industries with different financing needs and growth
opportunities. This underscores the importance of firm size, firm location, and industry finance
characteristics when assessing the impact of financial institutional reforms (Beck, DemirgucKunt, and Maskimovic (2005)). The results are robust to IV analysis, to controlling for other
institutional factors that can explain cross-country variation in tax evasion, and to country fixed
effects that control for unobserved factors, thus highlighting the importance of financial sector
policies in addressing widespread tax evasion in many developing countries. Critically, our
findings are robust to controlling for a standard measure of financial depth, suggesting that
specific outreach dimensions have a first-order effect on real sector outcomes. Finally, our
findings are confirmed in a smaller panel sample of surveys where we can exploit within-firm
variation. We show that the same firms report lower tax evasion after the introduction of or
improvements in credit information sharing and expansion in bank branch penetration. Our
findings suggest a financial system that provides easier access to credit increases opportunity
costs of tax evasion. They also show that financial sector outreach is an important policy lever to
bring more small firms into the formal economy.
These results are novel in the literature. While previous papers show the explanatory
power of financial sector development for variation in tax evasion across countries, this paper is
the first to show the explicit link between two dimensions of financial intermediary development
and cross-firm and cross-country variation in tax evasion. While previous papers focus on the
34
relationship between governance conflicts within firms and tax avoidance, using data mostly
from the U.S., this is the first paper to relate cross-country variation in information asymmetries
and agency problems between financial institutions and borrowers with tax evasion.
We see this paper as a first exploration of the relationship between financial sector
outreach and tax evasion. As more data become available, other aspects of financial sector
outreach can be linked to tax evasion.
35
Appendix
Variable Definitions and Data Sources
Variable
Definition
Original sources
Tax evasion ratio
Question c241: Recognizing the difficulties many enterprises face in fully
complying with taxes and regulations, what percentage of total sales would you
estimate the typical establishment in your area of activity reports for tax
purposes? The tax evasion ratio is equal to one minus the answered number.
World Bank Private
Enterprise Survey
Tax evasion dummy
Equals one if tax evasion ratio is greater than zero, otherwise zero.
Firm-level data
Firm location
Question c2071: Where are this establishment and your headquarters located in
this country?
(Enumerator, Please code as follows: 1=capital city; 2=other city of over 1
million population; 3=city of 250,000-1million; 4=city of 50,000-250,000;
5=town or location with less than 50,000 population).
Capital city
Firm location = 1 (capital city).
Small city
Firm location = 4 and 5 (city of 50,000-250,000 and town or location with less
than 50,000 population)
Employment
Total employment of the firm.
Small firm
Large firm
World Bank Private Enterprise Survey definition: those firms with less than 20
employees
World Bank Private Enterprise Survey definition: those firms with 100 and over
employees.
Foreign
Proportion of the firm is owned by foreign investors (Question c203b).
Exporter
Export dummy equal one if the firm exports, otherwise zero.
Firm age
Firm auditing
Calculated from Question c201: In what year did your firm begin operations in
this country?
Dummy equal one if financial statements of the firm are reviewed by an
external auditor (Question c232).
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
World Bank Private
Enterprise Survey
Country-level data
Demo branch
Demographic branch penetration: number of bank branches per 1,000,000
adults.
Geo branch
Geographic branch penetration: number of bank branches per 10,000 sq km.
Depth of credit information
An index that measures the information content of the credit information. A
value of one is added to the index when a country’s information agencies have
each of these characteristics: (1) both positive credit information (for example,
loan amounts and pattern of on-time repayments) and negative information (for
example, late payments, number and amount of defaults and bankruptcies) are
distributed; (2) data on both firms and individual borrowers are distributed; (3)
data from retailers, trade creditors, or utilities, as well as from financial
institutions, are distributed; (4) more than two years of historical data are
distributed; (5) data are collected on all loans of value above 1% of income per
capita; and (6) laws provide for borrowers’ right to inspect their own data. The
index ranges from zero to six, with higher values indicating the availability of
more credit information, from either a public registry or a private bureau, to
facilitate lending decisions.
36
IMF: fas.imf.org,
Beck et al. (2007),
and Kendall et al.
(2010)
IMF: fas.imf.org,
Beck et al. (2007),
and Kendall et al.
(2010)
Djankov et al.
(2007),
World Bank “Doing
Business” database
Information sharing dummy
The dummy variable equals one if an information sharing agency (public
registry or private bureau) operates in the country, zero otherwise.
Public credit registry
A dummy variable that equals one if a public registry operates in the country
during the sample period, zero otherwise.
Private bureau
A dummy variable that equals one if a private bureau operates in the country
during the sample period, zero otherwise.
Public credit registry coverage
An indicator that reports the number of individuals and firms listed in a public
credit registry with current information on repayment history, unpaid debts, or
credit outstanding. The number is expressed as a percentage of the adult
population. A public credit registry is defined as a database managed by the
public sector, usually by the central bank or the superintendent of banks, that
collects information on the creditworthiness of borrowers (persons or
businesses) in the financial system and makes it available to financial
institutions. If no public registry operates, the coverage value is zero.
Djankov et al.
(2007),
World Bank “Doing
Business” database
Djankov et al.
(2007),
World Bank “Doing
Business” database
Djankov et al.
(2007),
World Bank “Doing
Business” database
Djankov et al.
(2007),
World Bank “Doing
Business” database
Private credit bureau coverage
The private credit bureau coverage indicator reports the number of individuals
and firms listed by a private credit bureau with current information on
repayment history, unpaid debts or credit outstanding. The number is expressed
as a percentage of the adult population. A private credit bureau is defined as a
private firm or nonprofit organization that maintains a database on the
creditworthiness of borrowers (persons or businesses) in the financial system
and facilitates the exchange of credit information among banks and financial
institutions. Credit investigative bureaus and credit reporting firms that do not
directly facilitate information exchange among banks and other financial
institutions are not considered. If no private bureau operates, the coverage value
is 0.
Total tax rate
Total tax rate (proportion of commercial profits).
VAT dummy
A dummy variable that equals one if a country has VAT.
Log of time to prepare and pay
taxes (hours)
Log of time to prepare and pay taxes (hours).
Total number of taxes paid (log)
Log of the total number of taxes paid by businesses, including electronic filing.
Private credit / GDP
Ratio of private credit outstanding to GDP.
Beck et al. (2010)
Bank concentration (HHI)
To control for competition we use a Herfindahl index, defined as the sum of the
squared shares of bank assets to total assets within a given country.
BankScope
Crime
Log of per-100,000 population total crime rates.
United Nations
Office on Drugs and
Crime (UNODC)
Rule of law
An indicator that measures the extent to which agents have confidence in and
abide by the rules of society, and in particular the quality of contract
enforcement, the police, and the courts, as well as the likelihood of crime and
violence. Higher values mean stronger law and order.
Kaufmann et al.
(2010)
Control of corruption
Log GDP per capita (USD)
An indicator that measures the extent to which public power is exercised for
private gain, including both petty and grand forms of corruption, as well as
“capture” of the state by elites and private interests. Higher values indicate
better control of corruption.
Logarithm of gross domestic product per capita in US dollars.
37
Djankov et al.
(2007),
World Bank “Doing
Business” database
World Development
Indicators (WDI)
International VAT
Services: www.tmfvat.com/ vat
World Development
Indicators (WDI)
World Development
Indicators (WDI)
Kaufmann et al.
(2010)
World Development
Indicators (WDI)
IMF capital account
liberalization indicator
Equity market liberalization
intensity
Creditor rights
Following Bekaert et al. (2005) and Bekaert et al.(2007), we measure capital
account liberalization by employing the IMF’s Annual Report on Exchange
Arrangements and Exchange Restrictions (AREAER). The indicator takes on a
value of zero if the country has at least one control in the “restrictions on
payments for the capital account transaction” category, otherwise its value is
one.
The ratio of the Standard & Poor’s/International Finance Corporation investable
market capitalization (S&P/IFCI) to global market capitalization (S&P/IFCG).
The global market capitalization is intended to represent the overall market
portfolio for each country, whereas the investable market capitalization is
designed to represent a portfolio of domestic equities that are available to
foreign investors. Fully liberalized countries for which all of the stocks are
available to foreign investors have an intensity measure of one, and fully
segmented countries have an intensity measure of zero (see Bekaert et al. (2005)
and Bekaert et al.(2007)).
An index that measures the power of secured lenders in bankruptcy. A score of
one is assigned when each of the following rights of secured lenders is defined
in laws and regulations. First, there are restrictions, such as creditor consent, for
a debtor to file reorganization. Second, secured creditors are able to seize their
collateral after the reorganization petition is approved. Third, secured creditors
are paid first out of the proceeds of liquidating a bankrupt firm. Last,
management does not retain administration of its property pending the
resolution of the reorganization. The index ranges from zero to four. Higher
values indicate stronger creditor rights.
Property rights
Countries with more secure property rights and legal institutions that were more
supportive of rule of law receive higher ratings.
Voice and accountability
An indicator that measures the extent to which a country’s citizens are able to
participate in selecting their government, as well as freedom of expression,
freedom of association, and free media. Higher values mean greater political
rights.
Government effectiveness
Political stability
Quality of regulation
The indicator measures the quality of public services, the quality of the civil
service and the degree of its independence from political pressures, the quality
of policy formulation and implementation, and the credibility of the
government’s commitment to such policies. Higher values mean higher quality
public and civil service.
The indicator measures the perceptions of the likelihood that the government
will be destabilized or overthrown by unconstitutional or violent means,
including political violence and terrorism. Higher values mean more stable
political environment.
An indicator that measures the ability of the government to formulate and
implement sound policies and regulations that permit and promote market
competition and private sector development. Higher values mean higher quality
regulation.
Bekaert et al.
(2005), Bekaert et
al.(2007),
IMF
Bekaert et al.
(2005), Bekaert et
al.(2007),
Datastream,
Standard & Poor’s
Global Stock Market
Factbooks
Djankov et al.
(2007)
Fraser Institute
Website
Kaufmann et al.
(2010)
Kaufmann et al.
(2010)
Kaufmann et al.
(2010)
Kaufmann et al.
(2010)
Industrial-level data
EFD (external finance
dependence)
GO1 (industrial growth
opportunities measure)
GO2 (global growth
opportunities at the country
level)
The fraction of capital expenditures not financed with internal funds for both
public and private U.S. firms in the same industry over 2002 to 2006. It is based
the approach of Rajan and Zingales (1998).
U.S. industrial valued-added growth rate over 1990 to 1999.
Following Bekaert et al. (2007), an annual measure constructed as the industry
composition for each country by output share according to UNIDO Industrial
Statistics Database. The price-earnings (PE) ratio for each industry at the global
38
Capital IQ
UNIDO Industrial
Statistics Database
Bekaert et al.
(2007), Datastream,
and UNIDO
Industrial Statistics
level is used to construct an implied measure of growth opportunities for each
country by weighting each global industry PE ratio by its relative share for that
country. This measure then is subtracted from the overall world market PE ratio
to remove world discount rate effects (and is also subtracted from a five-year
moving average). The difference is “growth opportunities” (LGO_MA), that is,
LGO_MAi,t = LGOi,t - Σs=1 to 5 LGOi,t-s, where LGOi,t = ln[(IPEt Wi,t)/(IPEt Wt)],
IPEt is a vector of global industry PE ratios, Wi,t is a vector of country-specific
industry weights, and Wt is a vector of world industry weights.
Database
Instrumental variables
Proportion of other countries in
the same region that have credit
registries
Proportion of other countries in the same region (such as Asia, Europe, etc.) that
have credit registries.
Bank supervisor tenure (years)
The average tenure of current supervisors.
log of # of bank supervisors
Log of the number of bank supervisors.
Independence of supervisory
authority - overall
Official supervisory power
Foreign bank ownership
Private bank ownership
Nonperforming loan
The degree to which the supervisory authority is independent of the government
and legally protected from the banking industry. The indicator is constructed
based on the following three questions. (1) Are the supervisory bodies
responsible to a) the Prime Minister, b) the Finance Minister or other senior
government officials, or c) a legislative body (yes=1)? (2) Can the supervisors
be sued if they take actions against a bank (No=1)? (3) Does the chair of the
supervisory agency have a fixed term contract and how long? (=1 if term>=4).
A higher value means a more independent supervisory agency.
An index aggregating supervisory power. Specifically, it indicates whether the
supervisory agency has the legal right to meet directly with external auditors to
discuss their report without getting approval from the bank; receive direct report
from the external auditor on any presumed involvement of bank management in
various types of misconduct; take actions against external auditors for
negligence; change a bank’s internal organizational structure; get access to
information on off-balance-sheet items; require the bank management to
constitute provisions to cover actual or potential losses; suspend the board’s
decision to distribute dividends, bonuses, and management fees; declare a
bank’s insolvency; intervene in the ownership rights in a problem bank;
supersede shareholder rights; and replace management and directors.
The fraction of the banking system's assets in banks that are 50% or more
owned by foreigners.
The fraction of the banking system's assets in banks that are 50% or more
owned privately.
The previous five-year average ratio of nonperforming loans to total loans.
39
Djankov et al.
(2007),
World Bank “Doing
Business” database
Barth, Caprio, and
Levine (2006, 2008)
Barth, Caprio, and
Levine (2006, 2008)
Barth, Caprio, and
Levine (2006, 2008)
Barth, Caprio, and
Levine (2006, 2008)
Barth, Caprio, and
Levine (2006, 2008)
Barth, Caprio, and
Levine (2006, 2008)
Bankscope
REFERENCES
Acemoglu, Daron, and Simon Johnson, 2005, Unbundling institutions, Journal of Political
Economy 113, 949-995.
Ai, Chunrong, and Edward Norton, 2003, Interaction terms in logit and probit models,
Economics Letters 80, 123−129.
Alexeev, Michael, Eckhard Janeba, and Stefan Osborne, 2005, Taxation and evasion in the
presence of extortion by organized crime, Journal of Comparative Economics 32, 375-387.
Ayyagari, Meghana, Asli Demirguc-Kunt, and Vojislav Maksimovic, 2008, How well do
institutional theories explain firms’ perceptions of property rights? Review of Financial
Studies 21, 1833-1871.
Ayyagari, Meghana, Asli Demirguc-Kunt, and Vojislav Maksimovic, 2010, Formal versus
informal finance: Evidence from China, Review of Financial Studies 23, 3048-3097.
Ayyagari, Meghana, Asli Demirguc-Kunt, and Vojislav Maksimovic, 2013, Bribe payments and
innovation in developing countries: Are innovating firms disproportionately affected?
Journal of Financial and Quantitative Analysis, forthcoming.
Barth, James, Gerard Caprio, and Ross Levine, 2006, Rethinking Bank Regulation: Until Angels
Govern. Cambridge: Cambridge University Press.
Barth, James R., Gerard Caprio, and Ross Levine Ross, 2008, Bank regulations are changing:
For better or worse? Comparative Economic Studies 50, 537-563.
Barth, James, Chen Lin, Ping Lin, and Frank Song, 2009, Corruption in bank lending to firms:
Cross-country micro-evidence on the beneficial role of competition and information sharing,
Journal of Financial Economics 91, 361-388.
Beck, Thorsten, Asli Demirguc-Kunt, and Ross Levine, 2006, Bank supervision and corruption
in lending, Journal of Monetary Economics 53, 2131-2163.
Beck, Thorsten, Asli Demirguc-Kunt, and Vojislav Maksimovic, 2005, Financial and legal
constraints to firm growth: Does size matter? Journal of Finance 60, 137-177.
40
Beck, Thorsten, Asli Demirguc-Kunt, and Maria Soledad Martinez Peria, 2007, Reaching out:
Access to and use of banking services across countries, Journal of Financial Economics 85,
234-266.
Beck, Thorsten, Ross Levine, and Norman Loayza, 2000, Finance and the sources of growth,
Journal of Financial Economics 58, 261-300.
Bekaert, Geert, Campbell R. Harvey, Christian Lundblad, 2005, Does financial liberalization
spur growth? Journal of Financial Economics 77, 3-55.
Bekaert, Geert, Campbell R. Harvey, and Christian Lundblad, 2011, Financial openness and
productivity, World Development 39, 1-19.
Bekaert, Geert, Campbell R. Harvey, Christian Lundblad, and Stephan Siegel, 2007, Global
growth opportunities and market integration, Journal of Finance 62, 1081-1137.
Berger, Allen, Asli Demirguc-Kunt, Ross Levine, and Joseph Haubrich, 2004, Bank
concentration and competition: An evolution in the making, Journal of Money, Credit and
Banking 36, 433-451.
Brown, Martin, Tullio Jappelli, and Marco Pagano, 2009, Information sharing and credit: Firmlevel evidence from transition countries, Journal of Financial Intermediation 18, 151-172.
Caballero, Ricardo, and Arvind Krishnamurthy, 2004, Fiscal policy and financial depth, NBER
Working Paper No. 10532.
Caprio, Lorenzo, Mara Faccio, and John McConnell, 2013, Sheltering corporate assets from
political extraction, Journal of Law, Economics, and Organization, forthcoming.
Chen, Shuping, Xia Chen, Qjang Cheng, and Terry Shevlin, 2010, Are family firms more tax
aggressive than non-family firms? Journal of Financial Economics 95, 41-61.
Dabla-Norrisa, Era, Mark Gradstein, and Gabriela Inchaust, 2008, What causes firms to hide
output? The determinants of informality, Journal of Development Economics 85, 1-27.
Demirguc-Kunt, Asli, and Enrica Detragiache, 2002, Does deposit insurance increase banking
system stability? An empirical investigation, Journal of Monetary Economics 49, 1373-1406.
Demirguc-Kunt, Asli, and Vojislav Maksimovic, 1998, Law, finance and firm growth, Journal of
41
Finance 53, 2107-2137.
Desai, Mihir, and Dhammika Dharmapala, 2006, Corporate tax avoidance and high-powered
incentives, Journal of Financial Economics 79, 145-179.
Desai, Mihir, and Dhammika Dharmapala, 2009, Corporate tax avoidance and firm value,
Review of Economics and Statistics 91, 537-546.
Desai, Mihir, Alexander Dyck, and Luigi Zingales, 2007, Theft and taxes, Journal of Financial
Economics 84, 591-623.
Djankov, Simeon, Tim Ganser, Caralee McLiesh, Rita Ramalho, and Andrei Shleifer, 2010, The
effect of corporate taxes on investment and entrepreneurship, American Economic Journal:
Macroeconomics 2, 31-64.
Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, 2003,
Courts, Quarterly Journal of Economics 118, 457-522.
Djankov, Simeon, Caralee McLiesh, and Andrei Shleifer, 2007, Private credit in 129 countries,
Journal of Financial Economics 84, 299-329.
Durnev, Art, and Guriev, Sergei M., 2012, Expropriation risk and firm growth: A corporate
transparency channel, Working paper, University of Iowa.
Dyck, Alexander, and Luigi Zingales, 2004, Private benefits of control: An international
comparison, Journal of Finance 59, 537-600.
Farrell, Diana, 2004, The hidden dangers of the informal economy, The McKinsey Quarterly, 2637.
Fisman, Raymond, and Inessa Love, 2007, Financial dependence and growth revisited, Journal
of the European Economic Association 5, 470-479.
Friedman, Eric, Simon Johnson, Daniel Kaufmann, and Pablo Zoido-Lobaton, 2000, Dodging
the grabbing hand: The determinants of unofficial activity in 69 countries, Journal of Public
Economics 76, 459–493.
Gine, Xavier, Jessica Goldberg, and Dean Yang, 2012, Credit market consequences of improved
personal identification: Field experimental evidence from Malawi, American Economic
42
Review 102, 2923-2954.
Graham, John, Si Li, and Jiaping Qiu, 2008, Corporate misreporting and bank loan contracting,
Journal of Financial Economics 89, 44-61.
Gravelle, Jane, 2010, Tax havens: International tax avoidance and evasion, Congressional
Research Service Report for Congress.
Greene, William, 2004, Fixed effects and the incidental parameters problem in the Tobit model,
Econometric Review 23, 125-148.
Grossman, Herschel, 1995, Rival kleptocrats: The mafia vs. the state, in Gianluca Fiorentini and
Sam Pelzman, eds.: The Economics of Organized Crime (Cambridge University Press).
Hall, Robert, and Charles Jones, 1999, Why do some countries produce so much more output per
worker than others? Quarterly Journal of Economics 114, 83-116.
Hallward-Driemeier, Mary and Reyes Aterido, 2009, Comparing apples with ... apples: How to
make (more) sense of subjective rankings of constraints to business, World Bank Policy
Research Working Paper No. 5054.
Holmstrom, Bengt, and Jean Tirole, 1997, Financial intermediation, loanable funds, and the real
sector, Quarter Journal of Economics 112, 663-691.
Houston, Joel, Chen Lin, Ping Lin, and Yue Ma, 2010, Creditor rights, information sharing and
bank risk taking, Journal of Financial Economics 96, 485-512.
Houston, Joel, Chen Lin, and Yue Ma, 2012, Regulatory arbitrage and international bank flows,
Journal of Finance 67, 1845-1897.
IRS, 2007, General explanation of the administration’s fiscal year 2008 revenue proposals.
Jappelli, Tullio, and Marco Pagano, 2002, Information sharing, lending and defaults: Crosscountry evidence, Journal of Banking and Finance 26, 2017-2045.
Johnson, Simon, Daniel Kaufmann, John McMillan, and Christopher Woodruff, 2000, Why do
firms hide? Bribes and unofficial activity after communism, Journal of Public Economics 76,
495–520.
Johnson, Simon, Daniel Kaufmann, and Pablo Zoido-Lobaton, 1998, Regulatory discretion and
43
the unofficial economy, American Economic Review 88, 387-392.
Kashyap, Anil K, and Jeremy C. Stein, 1994, Monetary policy and bank lending, in Gregory N.
Mankiw, ed.: Monetary Policy (University of Chicago Press for NBER).
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi, 2010, The worldwide governance
indicators: A summary of methodology, data and analytical issues. World Bank Policy
Research Working Paper, 5430.
Kendall, Jake, Nataliya Mylenko, and Alejandro Ponce, 2010, Measuring Financial Access
around the world, World Bank Policy Research Working Paper No. 5253.
Kim, Jeong-Bim, Yinghua Li, and Liandong Zhang, 2011, Corporate tax avoidance and stock
price crash risk: Firm-level analysis, Journal of Financial Economics 100, 639-662.
King, Robert G., and Ross Levine, 1993, Finance and growth: Schumpeter might be right,
Quarterly Journal of Economics 108, 717–737.
Klapper, Leora, Luc Laeven, and Raghuram Rajan, 2006, Entry regulation as a barrier to
entrepreneurship, Journal of Financial Economics 82, 591-629.
Kroszner, Randall, Luc Laeven, and Daniela Klingebiel, 2007, Banking crises, financial
dependence, and growth, Journal of Financial Economics 84, 187–228.
La Porta, Rafael, and Andrei Shleifer, 2008, The unofficial economy and economic development,
Brookings Papers on Economic Activity, Fall, 275-352.
Levine, Ross, 2005, Finance and growth: Theory and evidence, in Phillippe Aghion and Stephen
Durlauf, eds: Handbook of Economic Growth, Vol. 1 (Elsevier).
Levine, Ross, and Sara Zervos, 1998, Stock markets, banks, and economic growth, American
Economic Review 88, 537-558.
Lin, Chen, Yue Ma, Paul Malatesta, and Yuhai Xuan, 2011, Ownership structure and the cost of
corporate borrowing, Journal of Financial Economics 100, 1-23.
Love, Inessa, 2003, Financial development and financing constraints: International evidence
from the structural investment model, Review of Financial Studies 16, 765-791.
Love, Inessa, and Nataliya Mylenko, 2003, Credit reporting and financing constraints, World
44
Bank Policy Research Working Paper No. WPS 3142.
Manski, Charles, and Steven Lerman, 1977, The estimation of choice probabilities from choice
based samples, Econometrica 45, 1977-1988.
Miller, Margaret, 2003, Credit reporting systems around the globe: The state of the art in public
credit registries and private credit reporting firms, in Margaret Miller, ed.: Credit Reporting
Systems and International Economy (MIT Press).
Padilla, Jorge, and Marco Pagano, 1997, Endogenous communication among lenders, and
entrepreneurial incentives, Review of Financial Studies 10, 205-236.
Pagano, Marco, and Tullio Jappelli, 1993, Information sharing in credit markets, Journal of
Finance 48, 1694-1718.
Pande, Rohini, and Christopher Udry, 2006, Institutions and development: A view from below,
in Richard Blundell, Whitney Newey, and Torsten Persson, eds.: Advances in Economics and
Econometrics (Cambridge University Press).
Petersen, Mitchell, 2009, Estimating standard errors in finance panel data sets: Comparing
approaches, Review of Financial Studies 22, 435-480.
Pitt, Mark, 2011, Overidentification tests and causality, Working Paper, Brown University
mimeo.
Rajan, Raghuram, and Luigi Zingales, 1998, Financial dependence and growth, American
Economic Review 88, 559-587.
Sarte, Pierre-Daniel, 2000, Informality and rent-seeking bureaucracies in a model of long-run
growth, Journal of Monetary Economics 46, 173-197.
Schneider, Friedrich, and Dominik Ernste, 2000, Shadow economies: Size, causes and
consequences, Journal of Economic Literature 38, 77-114.
Staiger, Douglas, and James H. Stock, 1997, Instrumental variables regressions with weak
instruments, Econometrica 65, 557-586.
Stiglitz, Joseph E., and Weiss Andrew, 1981, Credit rationing in markets with imperfect
information, The American economic review 71, 393-410.
45
Stulz, Rene, and Herb Johnson, 1985, An analysis of secured debt, Journal of Financial
Economics 14, 501-521.
Wooldridge, Jeffrey M., 2002, Econometric Analysis of Cross Section and Panel Data (MIT
Press).
46
Table I
Tax Evasion and Key Macro Variables Across Sample Countries
This table reports the mean of tax evasion and key macro variables across countries. Variable definitions are
provided in the Appendix.
1
Albania
Tax
evasion
ratio
(mean)
0.221
0.678
0.572
0.105
391
2
Algeria
0.254
0.706
0
0
0.481
0.045
0.122
85
3
Angola
0.584
0.744
1
3
0.530
0.040
0.107
555
4
Argentina
0.175
0.504
1
6
1.332
0.141
0.130
947
5
Armenia
0.066
0.298
0.656
1.968
1.125
0.942
0.076
500
6
Azerbaijan
0.137
0.367
0.639
2.556
0.624
0.463
0.081
493
7
Bangladesh
0.019
0.058
1
2
0.703
5.205
0.373
1,326
8
Belarus
0.075
0.267
1
3
0.574
0.231
0.126
547
9
Benin
0.140
0.406
1
1
0.513
0.128
0.146
155
0.202
0.475
1
6
0.586
0.031
0.378
552
0.209
0.411
1
5
1.732
1.081
0.381
299
12
Bolivia
Bosnia and
Herzegovina
Botswana
0.649
0.688
1
4
0.742
0.017
0.203
551
13
Brazil
0.327
0.830
1
5
2.034
0.305
0.287
1,502
14
Bulgaria
0.086
0.274
1
4.651
7.512
4.595
0.472
1,722
15
Burkina Faso
0.126
0.307
1
1
0.333
0.054
0.176
463
16
Burundi
0.157
0.415
1
1
0.163
0.280
0.242
270
17
Cambodia
0.516
0.908
0
0
0.229
0.113
0.072
414
18
Cape Verde
0.113
0.198
1
3
1.794
1.365
0.449
91
19
Chile
0.082
0.212
1
5
1.318
0.217
0.807
1,798
20
China
0.419
0.494
1
2
0.173
0.183
1.189
811
21
Colombia
Congo, Dem.
Rep.
Costa Rica
0.171
0.363
1
5
1.338
0.367
0.334
920
0.465
0.807
0
0
0.051
0.008
0.046
628
0.284
0.684
1
5
1.802
1.097
0.356
275
0.099
0.392
0
0
2.942
1.963
0.484
372
0.116
0.477
1
4.576
1.782
1.994
0.343
554
0.493
0.736
1
5
0.983
1.264
0.233
182
27
Croatia
Czech
Republic
Dominican
Republic
Ecuador
0.240
0.423
1
4.629
1.389
0.438
0.221
964
28
Egypt
0.156
0.335
1
3.737
0.420
0.217
0.472
3,000
29
El Salvador
0.207
0.436
1
5.680
0.732
1.458
0.428
815
30
Estonia
0.058
0.390
0.453
2.264
1.947
0.523
0.560
318
31
Ethiopia
0.389
0.624
1
2
0.094
0.040
0.238
484
Country
10
11
22
23
24
25
26
Tax
evasion
dummy
(mean)
0.685
Informati
on
sharing
(mean)
0
Depth of
credit
information
(mean)
0
47
Demo
branch
(mean)
Geo branch
(mean)
Private
credit / GDP
(mean)
Observ
ations
32
Gambia
0.681
0.892
0
0
0.492
0.420
0.156
166
33
Georgia
0.267
0.588
0
0
0.657
0.355
0.105
357
34
Germany
0.057
0.449
1
6
2.016
4.090
1.126
1,173
35
Ghana
0.271
0.557
1
0
0.448
0.272
0.145
494
36
Greece
0.109
0.527
1
4
3.467
2.558
0.796
478
37
Guatemala
0.250
0.546
1
5
1.806
1.209
0.271
942
38
Guyana
0.276
0.764
0
0
0.618
0.015
0.533
144
39
Honduras
0.234
0.501
1
4.008
1.028
0.391
0.415
720
40
Hungary
0.109
0.408
1
5
2.059
1.929
0.462
809
41
India
0.154
0.505
1
2
0.896
2.317
0.394
2,112
42
Indonesia
0.271
0.444
1
2
1.197
1.000
0.229
707
43
Ireland
0.038
0.289
1
5
3.457
1.658
1.606
464
44
Jordan
0.266
0.281
1
2
1.801
0.713
0.883
494
45
Kazakhstan
0.095
0.292
0
0
0.374
0.016
0.304
763
46
Kenya
0.199
0.490
1
2.990
0.325
0.119
0.263
875
47
0.099
0.442
1
5
1.741
7.002
0.870
498
0.205
0.506
0.444
0.887
0.492
0.089
0.060
435
49
South Korea
Kyrgyz
Republic
Lao PDR
0.039
0.154
1
0
0.158
0.025
0.058
279
50
Latvia
0.099
0.371
0.482
1.445
3.086
0.977
0.497
353
51
Lebanon
0.347
0.673
1
5
2.955
8.641
0.705
275
52
Lesotho
0.152
0.393
0
0
0.244
0.099
0.065
28
53
Liberia
0.320
0.900
1
1
0.291
0.065
0.160
150
54
0.131
0.435
1
4.312
1.317
0.593
0.270
375
0.264
0.585
1
3
1.698
1.086
0.209
316
56
Lithuania
Macedonia,
FYR
Madagascar
0.115
0.305
1
1
0.136
0.025
0.109
679
57
Malawi
0.288
0.539
0
0
0.331
0.140
0.079
115
58
Mali
0.219
0.500
1
1
0.659
0.020
0.183
732
59
Mauritania
0.470
0.828
1
1
0.399
0.007
0.206
227
60
Mauritius
0.193
0.286
0.684
2.052
1.981
9.495
0.804
497
61
Mexico
0.237
0.557
1
6
1.132
0.435
0.197
1,304
62
Moldova
0.169
0.518
0
0
0.758
0.669
0.210
560
63
Mongolia
0.368
0.779
1
3
3.992
0.045
0.256
149
64
Morocco
0.039
0.156
1
1
0.986
0.459
0.426
827
65
Mozambique
0.547
0.734
1
3
0.236
0.036
0.135
477
66
Namibia
0.254
0.370
1
5
0.767
0.012
0.485
322
67
Nicaragua
0.377
0.621
1
4.068
0.595
0.168
0.290
781
68
Nigeria
0.303
0.690
1
0
0.509
0.470
0.253
1,889
69
Oman
0.236
0.373
0
0
2.163
0.111
0.369
118
70
Pakistan
0.132
0.153
1
4
0.791
1.064
0.297
746
71
Panama
0.371
0.526
1
6
1.863
0.516
0.884
548
48
55
48
72
Paraguay
0.192
0.430
1
6
0.402
0.039
0.169
463
73
Peru
0.127
0.325
1
6
1.231
0.183
0.185
718
74
Philippines
0.218
0.583
1
3
1.257
2.140
0.331
542
75
Poland
0.100
0.415
1
4
1.997
2.087
0.284
1,462
76
Portugal
0.082
0.354
1
4
6.799
6.615
1.412
438
77
0.088
0.329
1
4.673
1.650
1.326
0.168
790
0.172
0.445
0
0
1.687
0.125
0.231
935
79
Romania
Russian
Federation
Rwanda
0.190
0.325
1
2
0.071
0.158
0.112
209
80
Saudi Arabia
0.907
0.958
1
5
0.775
0.058
0.369
621
81
Senegal
0.303
0.427
1
1
0.719
0.144
0.219
691
82
Serbia
0.233
0.656
1
0
0.565
0.392
0.193
331
83
Montenegro
0.171
0.425
1
2.365
2.277
0.849
0.242
433
84
0.158
0.773
0
0
0.266
0.121
0.095
150
0.081
0.362
1
3
1.988
1.830
0.370
326
86
Sierra Leone
Slovak
Republic
Slovenia
0.128
0.471
1
3
2.116
1.801
0.474
395
87
South Africa
0.149
0.303
1
5.626
0.711
0.187
1.458
1,494
88
Spain
0.037
0.185
1
5
10.006
7.438
1.457
579
89
Sri Lanka
0.077
0.432
1
4
0.854
1.998
0.306
324
90
Swaziland
0.579
0.743
1
5
0.650
0.227
0.213
292
91
Syria
0.354
0.581
0
0
0.365
0.249
0.196
508
92
Tajikistan
0.222
0.550
0
0
0.536
0.147
0.163
402
93
Tanzania
0.409
0.700
0
0
0.118
0.029
0.111
654
94
Turkey
0.362
0.683
1
5
1.265
0.807
0.204
2,068
95
Uganda
0.462
0.721
0
0
0.116
0.088
0.101
544
96
Ukraine
0.122
0.284
0
0
0.379
0.263
0.250
1,051
97
Uruguay
0.147
0.462
1
6
1.346
0.195
0.237
368
98
Uzbekistan
0.060
0.207
0
0
3.984
1.617
0.203
523
99
0.112
0.481
1
3
0.331
0.683
0.659
948
0.129
0.257
1
2
0.838
2.824
0.414
370
101
Vietnam
West Bank
and Gaza
Yemen
0.516
0.767
1
2
0.186
0.047
0.074
296
102
Zambia
0.160
0.536
0
0
0.280
0.021
0.063
151
Total
0.213
0.463
0.842
3.129
1.363
1.157
0.400
64,438
78
85
100
49
Table II
Summary Statistics
The table reports the mean, standard deviation, minimum, maximum and number of observations of the key
variables. Variable definitions are provided in the Appendix.
Variables
Mean Std. Dev.
Min
Max
No. of countries Observations
Firm-level variables
Tax evasion dummy
0.463
0.499
0
1
102
64,438
Tax evasion ratio
0.213
0.315
0
1
102
64,438
Small firm dummy
0.478
0.500
0
1
102
64,438
Big firm dummy
0.208
0.406
0
1
102
64,438
Small city dummy
0.283
0.451
0
1
102
64,438
Capital city dummy
0.335
0.472
0
1
102
64,438
Log employment
3.335
1.545
0.693 10.365
102
64,438
Foreign
0.101
0.283
0
1
102
64,438
Exporter
0.203
0.402
0
1
102
64,438
Log firm age
2.652
0.863
0
5.568
102
64,438
Firm auditing dummy
0.490
0.500
0
1
102
64,438
Country-level variables
Information sharing dummy
Depth of credit information
Demo branch
Geo branch
Total tax rate
VAT dummy
Log of time to prepare and pay taxes
(hours)
Total number of taxes paid (log)
Private credit/GDP
Bank concentration (HHI)
Control of corruption
Rule of law
Crime
Government effectiveness
Log GDP per capita (USD)
IMF capital account liberalization
indicator
Equity market liberalization intensity
Industrial level variables
EFD (external finance dependence)
GO1 (industrial growth opportunities)
GO2 (global growth opportunities)
Instrumental variables
Proportion of other countries in the
same region that have credit
registries
Bank supervisor tenure (years)
Log of # of bank supervisors
Independence of supervisory
authority - overall
Official supervisory power
Foreign bank ownership
Private bank ownership
Nonperforming loan
0.842
3.129
1.363
1.157
0.489
0.891
0.365
2.139
1.689
1.798
0.180
0.312
0
0
0.044
0.006
0.132
0
1
6
10.006
10.049
0.967
1
102
102
102
102
102
102
5.930
0.681
3.951
7.863
102
3.344
0.400
0.238
-0.284
-0.252
6.644
-0.153
7.688
0.694
0.338
0.176
0.703
0.801
1.211
0.676
1.217
1.946
0.029
0.026
-1.485
-1.623
3.642
-1.724
4.812
4.990
1.620
0.950
1.862
1.731
8.420
1.755
10.790
102
102
102
102
102
102
102
102
0.083
0.276
0
1
60
0.423
0.450
0
1
60
0.254
2.302
0.169
1.268
3.292
0.382
-2.297
-7.000
-0.649
5.227
8.904
1.946
102
102
83
0.762
0.118
0.538
1
102
7.326
4.702
4.251
1.267
1
1.609
22.5
8.101
92
90
1.549
0.922
0
3
87
11.214
0.339
0.423
0.117
2.617
0.276
0.274
0.119
4
0
0
0.008
16
1
0.934
0.919
92
89
89
101
50
No. of
industries
26
26
26
Table III
Basic Results: Information Sharing, Financial Outreach, and Tax Evasion
For the Probit model in Panel A, the dependent variable is tax evasion dummy. For the Tobit model in Panel B, the
dependent variable is tax evasion ratio. The pooled sample period is 2002 to 2010. The estimation is based on crosssectional data and includes a full set of industry and year dummies. The omitted variables are medium-sized city,
domestic firms, and non-exporters. The marginal effects (dy/dx) of the regressions are presented. The marginal
effect of a dummy variable is calculated as the discrete change in the expected value of the dependent variable as the
dummy variable changes from zero to one. P-values are computed by the heteroskedasticity-robust standard errors
clustered for countries and are presented in brackets. *, **, and *** represent statistical significance at the 10%, 5%,
and 1% level, respectively. Variable definitions are provided in the Appendix.
Depth of credit information
Demo branch
Firm-level controls
Small city
Capital city
Log employment
Foreign
Exporter
Log firm age
Firm auditing
(1)
(2)
(3)
Panel A: Probit regressions
-0.078
-0.072
-0.076
[0.000]*** [0.000]*** [0.000]***
-0.073
-0.067
-0.064
[0.000]*** [0.002]*** [0.000]***
(4)
(5)
(6)
Panel B: Tobit regressions
-0.059
-0.059
-0.058
[0.000]*** [0.000]*** [0.003]***
-0.053
-0.052
-0.051
[0.004]*** [0.001]*** [0.000]***
0.061
[0.030]**
-0.057
[0.012]**
-0.040
[0.017]**
-0.075
[0.023]**
-0.022
[0.154]
-0.025
[0.211]
-0.039
[0.019]**
0.056
[0.032]**
-0.047
[0.124]
-0.042
[0.031]**
-0.079
[0.114]
-0.038
[0.059]*
-0.037
[0.072]*
-0.043
[0.080]*
Country-level controls related to tax system
Total tax rate
VAT dummy
Log of time to prepare and pay taxes (hours)
Total number of taxes paid (log)
0.042
[0.058]*
-0.054
[0.059]*
-0.039
[0.019]**
-0.066
[0.113]
-0.025
[0.153]
-0.023
[0.213]
-0.038
[0.018]**
0.038
[0.035]**
-0.051
[0.056]*
-0.037
[0.025]**
-0.076
[0.019]**
-0.027
[0.118]
-0.022
[0.287]
-0.033
[0.025]**
0.524
[0.015]**
-0.082
[0.079]*
0.019
[0.537]
0.136
[0.041]**
0.443
[0.017]**
-0.062
[0.117]
0.018
[0.482]
0.151
[0.038]**
Other country-level controls
Private credit/GDP
Log GDP per capita (USD)
-0.014
[0.147]
-0.013
[0.142]
-0.161
[0.057]*
0.210
[0.139]
0.024
[0.030]**
-0.057
[0.012]**
-0.058
[0.043]**
-0.068
[0.039]**
-0.010
[0.160]
Observations
Countries
Pseudo R2
64,438
102
0.132
64,438
102
0.140
64,438
102
0.145
Bank concentration (HHI)
Crime
Rule of law
Control of corruption
Government effectiveness
51
0.053
[0.062]*
-0.045
[0.066]*
-0.041
[0.032]**
-0.086
[0.071]*
-0.042
[0.039]**
-0.034
[0.187]
-0.041
[0.075]*
0.052
[0.028]**
-0.041
[0.012]**
-0.039
[0.029]**
-0.082
[0.115]
-0.044
[0.031]**
-0.031
[0.256]
-0.036
[0.066]*
0.292
[0.036]**
-0.089
[0.073]*
0.030
[0.040]**
0.054
[0.066]*
0.310
[0.034]**
-0.119
[0.032]**
0.022
[0.185]
0.056
[0.058]*
-0.032
[0.041]**
-0.020
[0.076]*
-0.118
[0.025]**
0.252
[0.026]**
0.037
[0.204]
-0.121
[0.015]**
-0.073
[0.052]*
-0.152
[0.213]
-0.025
[0.124]
64,438
102
0.153
64,438
102
0.162
64,438
102
0.179
Table IV
Instrumental Variables Tobit Estimation Second-Stage Regression Results
This table reports second-stage regression results of the IV Tobit estimation. The dependent variable is tax evasion
ratio. The endogenous variables are depth of credit information and demographic branch. The instrumental variables
are the proportion of other countries in the same region that have credit registries, bank supervisor tenure (years),
log of # of professional bank supervisors, independence of supervisory authority – overall, official supervisory
power, foreign bank ownership, private bank ownership, and previous five-year average ratio of non-performing
loans to total loans. Variable definitions are provided in the Appendix. The pooled sample periods are 2002 to 2010
for Panel A and 2006 to 2010 for Panel B. For Panel B, the sampling weights are the inverse of the probability of
selection of the firms surveyed. The estimation is based on cross-sectional data and includes a full set of industry
and year dummies. The omitted variables are medium-sized city, domestic firms, and non-exporters. The marginal
effects (dy/dx) of the regressions are presented. The marginal effect of a dummy variable is calculated as the discrete
change in the expected value of the dependent variable as the dummy variable changes from zero to one. P-values
are computed by heteroskedasticity-robust standard errors clustered by country and are presented in brackets. *, **,
and *** represent statistical significance at the 10%, 5%, and 1% level, respectively.
(1)
(2)
(3)
(4)
(5)
(6)
Panel A: Without sampling weights
Panel B: With sampling weights
Depth of credit information
-0.091
-0.089
-0.081
-0.109
-0.107
-0.103
[0.000]*** [0.002]*** [0.004]***
[0.000]*** [0.003]*** [0.000]***
Demo branch
-0.067
-0.063
-0.061
-0.071
-0.069
-0.069
[0.002]*** [0.000]*** [0.007]***
[0.002]*** [0.000]*** [0.006]***
Firm-level controls
Small city
0.063
0.065
0.062
0.061
0.054
0.053
[0.036]**
[0.037]**
[0.040]**
[0.026]**
[0.061]*
[0.035]**
Capital city
-0.052
-0.047
-0.041
-0.046
-0.049
-0.045
[0.065]*
[0.071]*
[0.042]**
[0.021]**
[0.020]**
[0.022]**
Log employment
-0.040
-0.043
-0.042
-0.023
-0.022
-0.020
[0.015]**
[0.011]**
[0.012]**
[0.054]*
[0.028]**
[0.034]**
Foreign
-0.080
-0.070
-0.076
-0.088
-0.082
-0.074
[0.036]**
[0.086]*
[0.076]*
[0.016]**
[0.020]**
[0.052]*
Exporter
-0.039
-0.044
-0.040
-0.127
-0.129
-0.121
[0.125]
[0.094]*
[0.092]*
[0.051]*
[0.056]*
[0.045]**
Log firm age
-0.045
-0.044
-0.039
-0.012
-0.013
-0.011
[0.091]*
[0.094]*
[0.118]
[0.274]
[0.305]
[0.276]
Firm auditing
-0.045
-0.046
-0.045
-0.052
-0.058
-0.059
[0.016]**
[0.015]**
[0.014]**
[0.023]**
[0.020]**
[0.019]**
Country-level controls related to tax system
Total tax rate
0.170
0.212
0.284
0.321
[0.022]**
[0.018]**
[0.017]**
[0.019]**
VAT dummy
-0.122
-0.116
-0.098
-0.093
[0.041]**
[0.061]*
[0.081]*
[0.154]
Log of time to prepare and pay taxes (hours)
0.016
0.011
0.008
0.009
[0.081]*
[0.272]
[0.296]
[0.279]
Total number of taxes paid (log)
0.030
0.028
0.020
0.024
[0.076]*
[0.072]*
[0.158]
[0.036]**
Other country-level controls
Private credit/GDP
-0.086
-0.151
[0.064]*
[0.039]**
Bank concentration (HHI)
0.191
0.153
[0.167]
[0.218]
Crime
0.031
0.030
52
Log GDP per capita (USD)
-0.010
[0.157]
-0.009
[0.210]
[0.067]*
-0.054
[0.020]**
-0.084
[0.331]
-0.118
[0.074]*
-0.008
[0.205]
Observations
Countries
Hansen’s overidentification test (p-value)
Pseudo R2
57,094
83
0.289
0.176
57,094
83
0.319
0.183
57,094
83
0.307
0.188
Rule of law
Control of corruption
Government effectiveness
53
-0.027
[0.016]**
-0.028
[0.018]**
[0.089]*
-0.117
[0.019]**
-0.156
[0.045]**
-0.129
[0.137]
-0.029
[0.017]**
21,541
34
0.224
0.168
21,541
34
0.229
0.176
21,541
34
0.280
0.184
Table V
Firm Location and Tax Evasion
The dependent variable is tax evasion ratio. Columns (1), (2), (5), and (6) are estimated by Tobit regressions.
Columns (3) and (4) are estimated by IV Tobit regressions where the endogenous variables are depth of credit
information and demographic branch, and their interactions with small city and capital city dummies. The
instrumental variables are the proportion of other countries in the same region that have credit registries, bank
supervisor tenure (years), log of # of professional bank supervisors, independence of supervisory authority – overall,
official supervisory power, foreign bank ownership, private bank ownership, the previous five-year average ratio of
non-performing loans to total loans, and their interactions with small city and capital city dummies (see. Wooldridge
(2002), p.234). Variable definitions are provided in the Appendix. The pooled sample period is 2002 to 2010. The
estimation is based on cross-sectional data and includes a full set of industry dummies. Columns (1) to (4) also
include the same set of macro controls (country-level controls related to the tax system and other country-level
controls) as in Table III. The omitted variables are medium-sized city, domestic firms, and non-exporters. The
marginal effects (dy/dx) of the regressions are presented. The marginal effect of a dummy variable is calculated as
the discrete change in the expected value of the dependent variable as the dummy variable changes from zero to one.
P-values are computed by heteroskedasticity-robust standard errors clustered for countries and are presented in
brackets. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level, respectively.
(1)
(2)
(3)
Tobit
Depth of credit information
Demo branch
Small city x Depth of credit information
Capital city x Depth of credit information
-0.060
[0.003]***
-0.061
[0.000]***
-0.026
[0.006]***
0.022
[0.027]**
-0.074
[0.000]***
-0.068
[0.014]**
-0.029
[0.007]***
0.025
[0.037]**
0.107
[0.022]**
-0.102
[0.028]**
-0.040
[0.041]**
-0.085
[0.016]**
-0.025
[0.143]
-0.043
[0.077]*
-0.035
[0.046]**
0.098
[0.012]**
-0.087
[0.063]*
-0.047
[0.014]**
-0.077
[0.197]
-0.047
[0.027]**
-0.039
[0.202]
-0.023
[0.247]
-0.076
[0.001]***
-0.063
[0.007]***
-0.026
[0.009]***
0.022
[0.035]**
-0.032
[0.021]**
0.014
[0.072]*
0.117
[0.016]**
-0.106
[0.058]*
-0.044
[0.016]**
-0.075
[0.217]
-0.045
[0.027]**
-0.036
[0.204]
-0.022
[0.248]
yes
yes
no
64,438
102
－
0.176
yes
yes
no
64,438
102
－
0.179
yes
yes
no
57,094
83
0.390
0.182
yes
yes
no
57,094
83
0.358
0.187
Capital city x Demo branch
Capital city
Log employment
Foreign
Exporter
Log firm age
Firm auditing
Macro controls
Year effects
Country effects x Year effects
Observations
Countries
Hansen’s over-identification test (p-value)
Pseudo R2
IV Tobit
-0.058
[0.002]***
-0.059
[0.000]***
-0.022
[0.014]**
0.019
[0.039]**
-0.033
[0.005]***
0.012
[0.023]**
0.113
[0.024]**
-0.109
[0.027]**
-0.035
[0.112]
-0.074
[0.123]
-0.024
[0.146]
-0.041
[0.079]*
-0.036
[0.052]*
Small city x Demo branch
Small city
(4)
54
(5)
(6)
Country x year
fixed-effects
-0.020
[0.014]**
0.015
[0.036]**
0.087
[0.031]**
-0.073
[0.062]*
-0.033
[0.081]*
-0.086
[0.011]**
-0.018
[0.108]
-0.024
[0.033]**
-0.017
[0.023]**
-0.018
[0.003]***
0.016
[0.053]*
-0.026
[0.026]**
0.012
[0.032]**
0.085
[0.032]**
-0.092
[0.054]*
-0.032
[0.115]
-0.087
[0.012]**
-0.016
[0.107]
-0.020
[0.065]*
-0.012
[0.064]*
no
－
yes
64,438
102
－
0.301
no
－
yes
64,438
102
－
0.315
Table VI
Firm Size and Tax Evasion
The dependent variable is tax evasion ratio. Columns (1), (2), (5), and (6) are estimated by Tobit regressions.
Columns (3) and (4) are estimated by IV Tobit regressions where the endogenous variables are depth of credit
information and demographic branch, and their interactions with small firm and big firm dummies. The instrumental
variables are the proportion of other countries in the same region that have credit registries, bank supervisor tenure
(years), log of # of professional bank supervisors, independence of supervisory authority – overall, official
supervisory power, foreign bank ownership, private bank ownership, the previous five-year average ratio of nonperforming loans to total loans, and their interactions with small firm and big firm dummies (see Wooldridge (2002),
p.234). Variable definitions are provided in the Appendix. The pooled sample period is 2002 to 2010. The
estimation is based on cross-sectional data and includes a full set of industry dummies. Columns (1) to (4) also
include the same set of macro controls (country-level controls related to the tax system and other country-level
controls) as in Table III. The omitted variables are medium-sized city, domestic firms, and non-exporters. The
marginal effects (dy/dx) of the regressions are presented. The marginal effect of a dummy variable is calculated as
the discrete change in the expected value of the dependent variable as the dummy variable changes from zero to one.
P-values are computed by the heteroskedasticity-robust standard errors clustered for countries and are presented in
brackets. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level, respectively.
(1)
(2)
(3)
Tobit
Depth of credit information
Demo branch
Small firm x Depth of credit information
Big firm x Depth of credit information
-0.058
[0.000]***
-0.068
[0.000]***
-0.021
[0.023]**
0.027
[0.034]**
0.078
[0.015]**
-0.062
[0.053]*
-0.047
[0.023]**
-0.066
[0.165]
-0.045
[0.027]**
-0.042
[0.269]
-0.036
[0.039]**
yes
yes
no
64,438
102
－
0.176
yes
yes
no
64,438
102
－
0.180
Big firm x Demo branch
Small city
Capital city
Log employment
Foreign
Exporter
Log firm age
Firm auditing
Macro controls
Year effects
Country effects x Year effects
Observations
Countries
Hansen’s overidentification test (p-value)
Pseudo R2
IV Tobit
-0.055
[0.007]***
-0.070
[0.003]***
-0.029
[0.015]**
0.023
[0.068]*
-0.014
[0.026]**
0.016
[0.018]**
0.071
[0.016]**
-0.051
[0.112]
-0.048
[0.024]**
-0.066
[0.164]
-0.046
[0.030]**
-0.041
[0.275]
-0.029
[0.054]*
Small firm x Demo branch
(4)
55
-0.075
[0.000]***
-0.071
[0.014]**
-0.022
[0.020]**
0.012
[0.084]*
(5)
(6)
Country x year
fixed-effects
0.063
[0.027]**
-0.052
[0.055]*
-0.050
[0.112]
-0.080
[0.040]**
-0.038
[0.127]
-0.038
[0.136]
-0.038
[0.184]
-0.079
[0.005]***
-0.072
[0.008]***
-0.024
[0.023]**
0.013
[0.086]*
-0.011
[0.020]**
0.019
[0.067]*
0.055
[0.030]**
-0.056
[0.121]
-0.046
[0.030]**
-0.072
[0.122]
-0.040
[0.126]
-0.039
[0.123]
-0.039
[0.171]
0.051
[0.030]**
-0.038
[0.061]*
-0.047
[0.035]**
-0.059
[0.197]
-0.024
[0.062]*
-0.018
[0.052]*
-0.018
[0.011]**
-0.018
[0.029]**
0.024
[0.035]**
-0.012
[0.034]**
0.016
[0.031]**
0.049
[0.031]**
-0.036
[0.064]*
-0.041
[0.054]*
-0.058
[0.195]
-0.023
[0.061]*
-0.019
[0.054]*
-0.017
[0.014]**
yes
yes
no
57,094
83
0.353
0.188
yes
yes
no
57,094
83
0.366
0.190
no
－
yes
64,438
102
－
0.331
no
－
yes
64,438
102
－
0.345
-0.016
[0.029]**
0.018
[0.069]*
Table VII
External Finance Dependence (EFD), Growth Opportunities (GO) and Tax Evasion
The dependent variable is tax evasion ratio. Columns (1), (2), (3), (7), (8), and (9) are estimated by Tobit regressions. Columns (4) to (6) are estimated by IV
Tobit regressions where the endogenous variables are depth of credit information and demographic branch, and their interactions with EFD, GO1, and GO2. The
instrumental variables are the proportion of other countries in the same region that have credit registries, bank supervisor tenure (years), log of # of professional
bank supervisors, independence of supervisory authority – overall, official supervisory power, foreign bank ownership, private bank ownership, previous fiveyear average ratio of non-performing loans to total loans, and their interactions with EFD, GO1, and GO2 for columns (4), (5), and (6), respectively (see
Wooldridge (2002), p.234). Variable definitions are provided in the Appendix. The pooled sample period is 2002 to 2010. The estimation is based on crosssectional data and includes a full set of industry dummies. Columns (1) to (6) also include the same set of macro controls (country-level controls related to tax
system and other country-level controls) as in Table III. The omitted variables are medium-sized city, domestic firms, and non-exporters. The marginal effects
(dy/dx) of the regressions are presented. The marginal effect of a dummy variable is calculated as the discrete change in the expected value of the dependent
variable as the dummy variable changes from zero to one. P-values are computed by heteroskedasticity-robust standard errors clustered for countries and are
presented in brackets. *, **, and *** represent statistical significance at the 10%, 5%, and 1% level, respectively.
(1)
Depth of credit information
Demo branch
EFD x Depth of credit information
EFD x Demo branch
-0.036
[0.000]***
-0.055
[0.007]***
-0.010
[0.006]***
-0.013
[0.026]**
GO1 x Depth of credit information
(2)
Tobit
-0.047
[0.017]**
-0.067
[0.002]***
GO2 x Depth of credit information
GO2 x Demo branch
Capital city
Log employment
Foreign
Exporter
Log firm age
(4)
-0.049
[0.000]***
-0.064
[0.015]**
-0.057
[0.000]***
-0.076
[0.014]**
-0.016
[0.000]***
-0.018
[0.000]***
-0.005
[0.021]**
-0.007
[0.023]**
GO1 x Demo branch
Small city
(3)
0.079
[0.008]***
-0.056
[0.053]*
-0.046
[0.025]**
-0.065
[0.178]
-0.035
[0.157]
-0.050
0.084
[0.009]***
-0.052
[0.130]
-0.045
[0.017]**
-0.068
[0.178]
-0.045
[0.062]*
-0.043
(5)
IV Tobit
-0.055
[0.013]**
-0.078
[0.001]***
(6)
(7)
(8)
(9)
Country x year fixed-effects
-0.064
[0.007]***
-0.073
[0.003]***
-0.009
[0.020]**
-0.011
[0.015]**
-0.006
[0.067]*
-0.009
[0.027]**
-0.021
[0.055]*
-0.035
[0.025]**
0.071
[0.021]**
-0.041
[0.034]**
-0.041
[0.013]**
-0.083
[0.039]**
-0.043
[0.083]*
-0.044
0.057
[0.013]**
-0.042
[0.018]**
-0.032
[0.052]*
-0.077
[0.164]
-0.056
[0.034]**
-0.041
56
0.076
[0.016]**
-0.043
[0.032]**
-0.036
[0.051]*
-0.072
[0.031]**
-0.066
[0.011]**
-0.037
-0.003
[0.037]**
-0.007
[0.059]*
-0.024
[0.034]**
-0.026
[0.020]**
0.072
[0.014]**
-0.047
[0.068]*
-0.041
[0.017]**
-0.079
[0.072]*
-0.037
[0.118]
-0.058
0.031
[0.057]*
-0.038
[0.014]**
-0.039
[0.018]**
-0.057
[0.134]
-0.020
[0.109]
-0.023
0.028
[0.054]*
-0.036
[0.063]*
-0.038
[0.021]**
-0.060
[0.123]
-0.027
[0.025]**
-0.020
-0.014
[0.000]***
-0.015
[0.000]***
0.035
[0.029]**
-0.032
[0.026]**
-0.024
[0.064]*
-0.052
[0.203]
-0.018
[0.174]
-0.018
Firm auditing
[0.036]**
-0.036
[0.054]*
[0.081]*
-0.035
[0.063]*
[0.087]*
-0.045
[0.030]**
[0.215]
-0.050
[0.020]**
[0.242]
-0.079
[0.147]
[0.027]**
-0.055
[0.132]
[0.026]**
-0.042
[0.038]**
[0.157]
-0.036
[0.011]**
[0.161]
-0.039
[0.158]
Macro controls
Year effects
Country effects x Year effects
Observations
Countries
Hansen’s overidentification test (p-value)
Pseudo R2
yes
yes
no
64,438
102
－
0.176
yes
yes
no
64,438
102
－
0.179
yes
yes
no
57,806
83
－
0.181
yes
yes
no
57,094
83
0.301
0.209
yes
yes
no
57,094
83
0.275
0.194
yes
yes
no
52,413
71
0.204
0.226
no
－
yes
64,438
102
－
0.338
no
－
yes
64,438
102
－
0.347
no
－
yes
57,806
83
－
0.316
57
Table VIII
Panel Data Estimation Results: Depth of Credit Information and Tax Evasion
The dependent variable is tax evasion ratio. Columns (1) to (4) are estimated by OLS without firm fixed effects. Columns (5) to (8) are estimated by OLS with
firm fixed effects. Columns (9) to (12) are via IV estimations with firm fixed effects, where the endogenous variables include, in addition to demographic branch,
the depth of credit information and its interactions with small city and capital city dummies, small firm and big firm dummies, EFD, GO1, and GO2. All the
columns also include the same set of macro controls (country-level controls related to tax system and other country-level controls) as in Table III. The eight basic
instrumental variables are proportion of other countries in the same region that have credit registries, bank supervisor tenure (years), log of # of professional bank
supervisors, independence of supervisory authority, official supervisory power, foreign bank ownership, private bank ownership, and previous five-year average
ratio of non-performing loans to total loans. Additional IVs are the interaction terms of these eight basic IVs with small city and capital city dummies, and with
small firm and big firm dummies. Furthermore, extra IVs include the interaction terms of the above eight basic IVs with EFD, GO1, and GO2 for the
corresponding columns (see Wooldridge (2002), p.234). Columns (4), (8), and (12) are based on a sub-sample of 15 countries for which all macro controls had
changes during the two surveys of different years. Variable definitions are provided in the Appendix. The estimation also includes year fixed effects and is based
on firm-level unbalanced panel data over the period 2002 to 2010. The omitted variables are medium-sized city, domestic firms, and non-exporters. P-values are
computed by heteroskedasticity-robust standard errors clustered by country and are presented in brackets. *, **, and *** represent statistical significance at the
10%, 5%, and 1% level, respectively.
(1)
Depth of credit information
(2)
(3)
(4)
OLS without firm effects
-0.042
-0.044
-0.035
-0.049
[0.000]*** [0.003]*** [0.014]** [0.006]***
Firm location effects
Small city x Depth of credit information
-0.021
[0.013]**
Capital city x Depth of credit information 0.016
[0.062]*
Firm size effects
Small firm x Depth of credit information -0.036
[0.021]**
Big firm x Depth of credit information
0.026
[0.014]**
Financial characteristics
EFD x Depth of credit information
-0.010
[0.007]***
GO1 x Depth of credit information
Small city
Capital city
(6)
(7)
(8)
OLS with firm effects
-0.034
-0.037
-0.027
-0.042
[0.001]*** [0.001]*** [0.012]** [0.017]**
(10)
(11)
IV with firm effects
-0.044
-0.046
-0.038
[0.001]*** [0.000]*** [0.018]**
-0.037
[0.006]***
(12)
-0.021
[0.015]**
0.017
[0.068]*
-0.024
[0.000]***
0.019
[0.055]*
-0.015
[0.012]**
0.009
[0.201]
-0.018
[0.009]***
0.012
[0.043]**
-0.013
[0.021]**
0.013
[0.062]*
-0.020
[0.027]**
0.026
[0.059]*
-0.024
[0.014]**
0.016
[0.018]**
-0.026
[0.013]**
0.010
[0.157]
-0.030
[0.023]**
0.021
[0.039]**
-0.021
[0.019]**
0.035
[0.030]**
-0.041
[0.023]**
0.018
[0.195]
-0.042
[0.022]**
0.017
[0.190]
-0.035
[0.025]**
0.019
[0.159]
-0.026
[0.038]**
0.017
[0.192]
-0.032
[0.039]**
0.018
[0.269]
-0.019
[0.072]*
0.024
[0.038]**
-0.014
[0.151]
0.029
[0.031]**
-0.020
[0.035]**
0.022
[0.072]*
-0.023
[0.083]*
0.021
[0.043]**
-0.036
[0.039]**
0.021
[0.265]
-0.034
[0.037]**
0.035
[0.176]
-0.015
[0.001]***
-0.002
[0.033]**
-0.081
[0.000]***
0.075
[0.015]**
-0.060
-0.012
[0.002]***
-0.001
[0.071]*
-0.019
-0.029
[0.001]*** [0.000]***
-0.079
[0.001]***
0.074
[0.014]**
-0.059
(9)
-0.022
[0.016]**
0.018
[0.067]*
GO2 x Depth of credit information
Other controls
Demo branch
(5)
-0.085
[0.002]***
0.073
[0.012]**
-0.058
-0.071
[0.000]***
0.079
[0.000]***
-0.059
-0.003
[0.039]**
-0.009
-0.016
[0.007]*** [0.000]***
-0.074
-0.073
-0.062
[0.000]*** [0.000]*** [0.014]**
58
-0.070
[0.006]***
-0.014
-0.019
[0.008]*** [0.000]***
-0.098
-0.096
-0.076
-0.086
[0.000]*** [0.000]*** [0.004]*** [0.000]***
Foreign
Exporter
Log employment
Log firm age
Firm auditing
Macro control
Industry fixed effects
Firm fixed effects
Observations
Countries
Hansen’s overidentification test
(p-value)
Adjusted R2
[0.118]
-0.028
[0.042]**
-0.012
[0.117]
-0.065
[0.029]**
-0.020
[0.086]*
-0.061
[0.087]*
[0.112]
-0.016
[0.129]
-0.021
[0.046]**
-0.066
[0.028]**
-0.020
[0.087]*
-0.061
[0.084]*
[0.124]
-0.027
[0.065]*
-0.019
[0.067]*
-0.062
[0.143]
-0.022
[0.081]*
-0.069
[0.074]*
[0.021]**
-0.014
[0.287]
-0.020
[0.261]
-0.065
[0.036]**
-0.015
[0.117]
-0.059
[0.036]**
-0.057
[0.067]*
-0.020
[0.136]
-0.038
[0.215]
-0.058
[0.030]**
-0.020
[0.176]
-0.036
[0.139]
-0.058
[0.064]*
-0.019
[0.365]
-0.039
[0.151]
-0.054
[0.127]
-0.009
[0.427]
-0.044
[0.206]
-0.062
[0.029]**
-0.025
[0.029]**
-0.056
[0.043]**
-0.062
[0.025]**
-0.025
[0.226]
-0.055
[0.083]*
-0.057
[0.057]*
-0.023
[0.109]
-0.067
[0.109]
-0.055
[0.079]*
-0.009
[0.240]
-0.032
[0.345]
yes
yes
no
7,671
42
yes
yes
no
7,671
42
yes
yes
no
6,935
34
yes
yes
no
2,620
15
yes
－
yes
7,671
42
yes
－
yes
7,671
42
yes
－
yes
6,935
34
yes
－
yes
2,620
15
yes
－
yes
7,149
38
yes
－
yes
7,149
38
yes
－
yes
6,629
32
yes
－
yes
2,620
15
－
－
－
－
－
－
－
－
0.152
0.149
0.232
0.308
0.141
0.131
0.136
0.120
0.151
0.150
0.137
0.141
0.164
0.166
0.162
0.150
59
Table IX
Panel Data Estimation Results: Financial Sector Outreach and Tax Evasion
The dependent variable is tax evasion ratio. Columns (1) to (4) are estimated by OLS without firm fixed effects. Columns (5) to (8) are estimated by OLS with
firm fixed effects. Columns (9) to (12) are via IV estimations with firm fixed effects, where the endogenous variables include, in addition to depth of credit
information, the demographic branch, and its interactions with small city and capital city dummies, small firm and big firm dummies, EFD, GO1, and GO2. All
the columns also include the same set of macro controls (country-level controls related to the tax system and other country-level controls) as in Table III. The
eight basic instrumental variables are the proportion of other countries in the same region that have credit registries, bank supervisor tenure (years), log of # of
professional bank supervisors, independence of supervisory authority – overall, official supervisory power, foreign bank ownership, private bank ownership, and
previous five-year average ratio of non-performing loans to total loans. Additional IVs are the interaction terms of these eight basic IVs with small city and
capital city dummies, and with small firm and big firm dummies. Furthermore, extra IVs include the interaction terms of the above eight basic IVs with EFD,
GO1, and GO2 for the corresponding columns (see Wooldridge (2002) p.234). Columns (4), (8), and (12) are based on a sub-sample of 15 countries for which all
macro controls had changes during the two surveys of different years. Variable definitions are provided in the Appendix. The estimation also includes year fixed
effects and is based on firm-level unbalanced panel data over the period 2002 to 2010. The omitted variables are medium-sized city, domestic firms, and nonexporters. P-values are computed by heteroskedasticity-robust standard errors clustered by country and are presented in brackets. *, **, and *** represent
statistical significance at the 10%, 5%, and 1% level, respectively.
(1)
Depth of credit information
Demo branch
Firm location effects
Small city x Demo branch
Capital city x Demo branch
Firm size effects
Small firm x Demo branch
Big firm x Demo branch
Financial characteristics
EFD x Demo branch
(4)
(5)
-0.035
[0.000]***
-0.072
[0.000]***
(2)
(3)
OLS without firm effects
-0.031
-0.030
[0.012]**
[0.000]***
-0.075
-0.072
[0.000]*** [0.012]**
-0.035
[0.006]***
-0.080
[0.000]***
-0.026
[0.003]***
0.039
[0.023]**
-0.024
[0.004]***
0.041
[0.021]**
-0.025
[0.017]**
0.040
[0.024]**
-0.029
[0.017]**
0.030
[0.065]*
-0.028
[0.016]**
0.037
[0.024]**
-0.030
[0.008]***
0.038
[0.021]**
-0.034
[0.017]**
-0.074
[0.000]***
-0.023
[0.000]***
0.046
[0.015]**
-0.013
[0.004]***
0.012
[0.065]*
-0.015
[0.008]***
0.022
[0.073]*
-0.017
[0.007]***
0.016
[0.062]*
-0.027
[0.007]***
0.031
[0.145]
-0.013
[0.019]**
0.029
[0.082]*
-0.015
[0.013]**
0.027
[0.119]
-0.019
[0.012]**
0.026
[0.159]
-0.004
[0.022]**
0.065
[0.012]**
-0.052
-0.052
[0.001]***
-0.096
[0.011]**
(10)
(11)
IV with firm effects
-0.047
-0.031
[0.012]**
[0.017]**
-0.073
-0.077
[0.000]*** [0.012]**
-0.026
[0.014]**
-0.078
[0.014]**
-0.027
[0.005]***
0.024
[0.043]**
-0.046
[0.015]**
0.024
[0.015]**
-0.023
[0.006]***
0.040
[0.069]*
-0.022
[0.007]***
0.039
[0.031]**
-0.034
[0.029]**
0.039
[0.028]**
-0.024
[0.005]***
0.020
[0.042]**
-0.036
[0.020]**
0.049
[0.026]**
-0.023
[0.014]**
0.035
[0.122]
-0.023
[0.015]**
0.053
[0.074]*
-0.020
[0.023]**
0.045
[0.094]*
-0.050
[0.012]**
-0.034
[0.000]***
-0.017
[0.032]**
GO2 x Demo branch
Capital city
(9)
-0.021
[0.025]**
GO1 x Demo branch
Other controls
Small city
(8)
-0.021
[0.003]***
-0.065
[0.003]***
(6)
(7)
OLS with firm effects
-0.027
-0.033
[0.001]*** [0.001]***
-0.060
-0.061
[0.014]**
[0.015]**
0.067
[0.012]**
-0.058
-0.023
[0.028]**
-0.003
[0.063]*
-0.045
[0.013]**
-0.051
[0.002]***
0.069
[0.011]**
-0.053
0.070
[0.014]**
-0.058
-0.007
[0.022]**
-0.018
[0.014]**
60
(12)
-0.020
[0.000]***
Foreign
Exporter
Log employment
Log firm age
Firm auditing
Macro control
Industry fixed effects
Firm fixed effects
Observations
Countries
Hansen’s overidentification
test (p-value)
Adjusted R2
[0.127]
-0.021
[0.076]*
-0.014
[0.138]
-0.040
[0.059]*
-0.020
[0.085]*
-0.055
[0.075]*
[0.034]**
-0.018
[0.115]
-0.021
[0.033]**
-0.045
[0.013]**
-0.020
[0.087]*
-0.054
[0.074]*
[0.138]
-0.029
[0.026]**
-0.016
[0.145]
-0.043
[0.015]**
-0.021
[0.062]*
-0.064
[0.060]*
[0.067]*
-0.017
[0.162]
-0.019
[0.154]
-0.038
[0.217]
-0.014
[0.128]
-0.055
[0.031]**
-0.039
[0.016]**
-0.019
[0.163]
-0.042
[0.147]
-0.039
[0.042]**
-0.018
[0.217]
-0.039
[0.151]
-0.041
[0.056]*
-0.020
[0.060]*
-0.041
[0.152]
-0.036
[0.125]
-0.015
[0.196]
-0.033
[0.061]*
-0.042
[0.012]**
-0.023
[0.032]**
-0.058
[0.109]
-0.039
[0.057]*
-0.021
[0.130]
-0.061
[0.067]*
-0.042
[0.017]**
-0.023
[0.100]
-0.071
[0.044]**
-0.041
[0.076]*
-0.011
[0.369]
-0.048
[0.079]*
yes
yes
no
7,671
42
yes
yes
no
7,671
42
yes
yes
no
6,935
34
yes
yes
no
2,620
15
yes
－
yes
7,671
42
yes
－
yes
7,671
42
yes
－
yes
6,935
34
yes
－
yes
2,620
15
yes
－
yes
7,149
38
yes
－
yes
7,149
38
yes
－
yes
6,629
32
yes
－
yes
2,620
15
－
－
－
－
－
－
－
－
0.154
0.150
0.236
0.215
0.131
0.130
0.142
0.129
0.150
0.151
0.144
0.131
0.160
0.162
0.158
0.142
61
1
Most of this literature finds a positive relationship between finance and growth, robust to reverse causation
and omitted variables biases. See, for example, Levine and Zervos (1998), Demirguc-Kunt and Maksimovi (1998),
and Beck, Levine, and Loayza (2000). Bekaert, Harvey, and Lundblad (2011) have similar findings for the effect of
financial openness, while Bekaert, Harvey, and Lundblad (2005) find that financial liberalization leads to higher
economic growth. For a detailed review of the literature, see Levine (2005).
2
See Kashyap and Stein (1994) for a survey on a rich literature that relates transmission channels of monetary
policy to access to finance by firms and banks. Caballero and Krishnamurthy (2004) show that lack of financial
depth constrains fiscal policy.
3
Even in the U.S. the compliance rate is officially estimated at only 86% (IRS (2007)).
4
Surowiecki, July 11, 2011, “Doger Mania,” The New Yorker. According to Nikos Lekkas, the head of the
Greek tax inspectorate, “tax evasion in Greece has reached 12 to 15 per cent of the gross national product. That is
€40 to €45 billion per year. If we could recover even half of that, Greece would have solved the problem.” (June 8,
2012, The Telegraph).
5
Bhatti et al., 2012, USA Today, January 31
6
The reputation losses might also affect the firm’s investors, customers, and suppliers and change the terms of
trade on which they do business with the firm, thus reducing the present value of the firm’s future cash flows and
value (Graham, Li, and Qiu (2008)).
7
In many countries, tax authorities rank higher than secured creditors in bankruptcy proceedings. Moreover,
firms engaging in tax evasion often face financial penalties, which would also hurt firms’ profitability.
8
In fact, tax information is often collected by credit registries or private bureaus and shared among financial
institutions (Miller (2003)).
9
Better financial development such as credit information sharing mitigates the problems of adverse selection
and moral hazard and allows better discrimination between borrowers according to their creditworthiness (e.g.,
Stiglitz and Weiss (1981), Pagano and Jappelli (1993), Barth et al. (2009)). While the average cost of capital might
be lower in markets with better financial sector outreach, the reputation loss might result in a more profound effect
1
due to the better screening and information sharing networks. Thus, overall, the average cost of capital goes down,
while the variation might go up.
10
As Holmstrong and Tirole (1997, p. 665) point out, “Firms with low net worth have to turn to financial
intermediaries, who can reduce the demand for collateral by monitoring more intensively. Thus, monitoring is a
partial substitute for collateral.” As can be seen in the Internet Appendix, the presence of collateral is less of a
constraint for access to finance in economies with better financial development.
11
For the relative effect of financial sector depth on the growth of small vsersus large firms, see, for example,
Beck, Demirguc-Kunt, and Maksimovic (2005).
12
In this context, our paper also relates to Ayyagari, Demirguc-Kunt and Maksimovic (2013), who gauge the
extent to which tax evasion and bribing public officials constitute growth constraints for innovating firms.
13
Among others, Desai and Dharmapala (2006) show that tax evasion activities and managerial diversion are
determined together. They further show that higher-powered incentives (reduced agency costs) result in less tax
evasion. In the same vein, Desai and Dharmapala (2009) show that higher tax evasion only leads to higher stock
value among U.S. firms in high quality firms. Chen et al. (2010) find that family firms are less aggressive in tax
evasion than nonfamily firms as they might be more affected by negative price reaction from minority shareholders,
while Kim, Li, and Zhang (2011) argue that tax avoidance allows managerial rent extraction and therefore is
associated with higher firm-specific stock price crashes.
14
Broadly speaking, the paper is also related to the determinants of unofficial economic activities (for example,
Johnson et al. (1998, 2000), Friedman et al. (2000), Dabla-Norris, Gradstein, and Inchauste (2008)).
15
See www.enterpriseseurveys.org for more details. Similar surveys were previously conducted under the
leadership of the World Bank and other IFIs in Africa (Regional Project on Enterprise Development), the Central
and Eastern European transition economies (Business Environment and Enterprise Performance Surveys) in the
1990s, and world-wide in 2000 (World Business Environment Survey).
16
Among the many studies using firm-level surveys, see, for example, Johnson et al. (2000), Beck, Demirguc-
Kunt, and Maksimovic (2005), Djankov et al. (2003), Beck, Demirguc-Kunt, and Levine (2006) and Barth et al.
(2009).
2
17
The within-country standard deviation is calculated using the deviations from country averages, whereas
the between-country standard deviation is calculated from the country averages.
18
This indicator is based on expert assessment of how widespread tax evasion is in a country, ranging from
zero – common – to 10 – not common. Several studies compare firm-level responses related to the business
environment with data from other sources and find a high correlation (see, for example, Hallward-Driemeyer and
Aterido (2009)).
19
In robustness tests, we confirm our finding with an alternative indicator, Geographic Branch Penetration,
which is the number of bank branches per 10,000 square kilometer.
20
Beck, Demirguc-Kunt, and Martinez Peria (2007) and subsequent data collections also present data on the
number of loan accounts and the average loan balance to income per capita, but these data are available for a much
smaller set of countries.
21
In robustness tests, we gauge the sensitivity of our results to the use of alternative indicators of credit
information sharing, including (i) a dummy variable indicating the existence of a credit registry, (ii) dummy
variables for the existence of a public or private credit registry, and (iii) indicators of private or public credit registry
coverage, measured as the number of firms and individuals listed in registries relative to the adult population.
22
Ideally, we would like to have an indicator of actual distance from the economic center of the country, but
are restricted to using this location indicator as a proxy variable.
23
See Berger et al. (2004) for an overview of this literature. Bank concentration might thus influence tax
evasion through its impact on the opportunity costs of tax evasion.
24
In robustness tests, we also include the firm-level survey response to the question about whether taxation is
an obstacle for the operations and growth of the enterprise, with the responses varying between zero (no obstacle)
and four (very severe obstacle). Our results are robust to this alternative indicator of tax burden.
25
The Internet Appendix may be found in the online version of this article.
26
The typical measures for firm size are the firm’s sales revenue or total assets. There is a concern that firm
size might be endogenous because firm size might be limited by tax evasion levels. Therefore, in this study, we
measure firm size using the number of employees. Small firms are defined as firms with less than 20 employees,
3
while large firms are defined as firms with more than 100 employees. We acknowledge that this might only help
alleviate but not completely eliminate the endogeneity concern. A similar concern might also apply to the firm’s
location as firms might choose their location for unobserved reasons.
27
Following Rajan and Zingales (1998), the U.S. is not included in our sample. The calculation of industry
values is based on data from U.S. firms for which financial market frictions are considered to be relatively small and
should reflect mostly demand.
28
Since these industry characteristics are significantly correlated with each other, we do not include them at the
same time.
29
However, cross-sectional Tobit models do not have this kind of problem (see Wooldridge (2002), p. 538).
30
The marginal effects and elasticities are computed at the mean of all variables and there might be variation
across the distribution. We also test the robustness of the results to alternative measures such as the existence of an
information-sharing agency and geographic bank branch penetration. The results are highly consistent.
31
The Sargan test, which requires i.i.d. errors, is not appropriate because the sample weights make sample
errors not i.i.d. Under this case, the Hansen J test is the appropriate overidentification test (Pitt (2011)). In an early
work, Manski and Lerman (1977) demonstrate the importance of taking nonrandom sampling into account in the
estimates and tests.
32
In robustness tests, we show that our OLS results also hold in the sample without sampling weights.
33
The 21 countries we dropped are: Algeria, Bangladesh, Egypt, Gambia, Indonesia, Jordan, Kenya, Lebanon,
Mauritania, Morocco, Mozambique, Nigeria, Oman, Pakistan, Philippines, Russian Federation, Saudi Arabia, Syria,
Turkey, West Bank and Gaza, and Yemen,
34
This question is only available in the surveys of 53 countries.
35
As documented in the literature (e.g., Hall and Jones (1999), Caprio, Faccio and McConnell (2013), Durnev
and Guriev (2012)), government expropriation risks might affect corporate behaviors and outcomes. In countries
with higher risks of expropriation by the government, firms might have stronger incentives to hide profits from
government officials.
36
These countries are Croatia, Czech Republic, Estonia, Germany, Greece, Hungary, Ireland, South Korea,
4
Oman, Poland, Portugal, Saudi Arabia, Slovak Republic, Slovenia, and Spain.
37
Responses vary between zero (none) to four (severe).
38
Specifically, we use the responses to the following question: Recognizing the difficulties many enterprises
face in fully complying with taxes and regulations, what percentage of total sales would you estimate the typical
firm in your area of activity keeps “off the books”? 1=None at all, 2= 1-10%, 3= 11-20%, 4= 21-30%, 5= 31-40%,
6= 41-50%, and 7=More than 50%.
5